AI Roadmap Pro for Shreeji Precision Components (fictional)

This report was generated automatically by software using AI. No human reviewed it before delivery. It is general information only, not professional, legal, tax or financial advice. Verify prices, vendor terms and legal requirements before acting.

Shreeji Precision Components (fictional) · 2026-10-08 · AI Marga Roadmap Pro

Executive summary

Shreeji Precision Components is a Rajkot CNC machining unit of 51 to 200 people, supplying auto and pump parts to OEM suppliers with some export orders. The base tools are in place: Tally / TallyPrime for accounts, Excel and Google Sheets, Google Workspace and the WhatsApp Business app. But much of the daily work still runs on paper and phone, and AI use so far is individual staff trying ChatGPT or Gemini on their own.

The biggest opportunity is quotation speed. Your answers say an RFQ takes two to three days because drawings, costing sheets and old quotes sit in different Excel files and WhatsApp chats. The filing is the real fix; AI comes second. Put everything in one shared Google Drive folder, the RFQ Desk, track every enquiry in one RFQ Register sheet, and let Gemini find the nearest past quote and draft the quotation and cover email. The person who owns the costing sheet still approves every rate. One working day is a realistic target for repeat and similar parts (an estimate, not a promise); new or complex parts will take longer.

Second is the weekly view the owner asked for: orders, rejections and pending payments on one page in Google Data Studio (formerly Looker Studio), fed by one Weekly Numbers sheet. Third is collections: a receivables ageing export from Tally every Monday, with reminders drafted by Gemini and sent by the accounts person.

What to do first: in days 1–30, set up the RFQ Desk, start the Monday reminder routine, and move staff AI use onto company Google Workspace accounts under a one-page policy. All three use tools the business already has. The dashboard and a purchase bill extraction pilot follow in days 31–60. Reorder planning and Gujarati and Hindi shop-floor SOPs start in days 61–90.

What to avoid for now: do not buy a paid ERP plan, a WhatsApp API platform or a document-AI platform in the first 90 days. Your budget band would cover them, but none of them fixes scattered files. And do not put customer drawings, costing sheets, bank statements or PAN and GST data into free personal AI accounts.

Where your business stands today

This assessment uses only your questionnaire answers. Nothing was checked on site.

Strengths

• Accounts already run on Tally / TallyPrime, so sales invoices and outstanding receivables should already be digital. That is the raw material for reminders and the dashboard.
• Google Workspace is already in use. It gives company accounts, shared Drive folders, Sheets and Gemini without a new purchase. Which Gemini features you get depends on your edition.
• A moderate monthly budget and medium tech comfort mean the team can set up apps with some help and is not limited to free tools.
• The goals are specific and measurable: quotations within a day, and a weekly view of orders, rejections and pending payments.

Gaps

• There is no single place for quotation inputs. Until there is, any AI tool will search the same scattered files.
• Your answers do not say where rejections and rework are recorded, or whether stock is kept in Tally. If either is on paper, a short weekly entry is needed before a dashboard or reorder planning can work.
• Supervisors are comfortable in Gujarati and Hindi, while buyer paperwork is in English. Work instructions and entry sheets should match the reader's language.
• Staff AI accounts outside company control are the most immediate data risk. The data protection section and the policy template cover this.

Overall: ready to start with what you have. The first month is mostly organising files and fixing routines, not buying software.

Your top AI opportunities, ranked

Ranked by value for your business. The start phase is shown beside each one: quick wins 3 and 4 start early although they rank lower.

1. Turn customer RFQs into quotations faster (start: days 1–30)

Why it fits: This is the problem you described in most detail: two to three days per quotation, because drawings, costing sheets and old quotes sit in different Excel files and WhatsApp chats. Your goal is one day.

Day to day: Every RFQ, whether it comes by email or WhatsApp, is saved the same day in the RFQ Desk folder on Google Drive and logged in the RFQ Register. The register has seven stages: Received, Costing, Draft ready, Approved, Sent, Won, Lost. The person preparing the quote asks Gemini Notebook for the closest past quotations and opens the source it cites. The costing owner fills the costing sheet. Gemini then drafts the quotation text and cover email from those figures, and the costing owner approves before it goes out.

Tools: Google Workspace with Gemini and Gemini Notebook (formerly NotebookLM). TranZact, a cloud ERP for SME manufacturers, is a later option, on its free plan only.

Effort: Medium. Most of the work is collecting old quotes into one place.

Benefit (estimate): Repeat and similar parts quoted within one working day instead of two to three, and much less time spent hunting for old rates.

First steps:

1. Create the RFQ Desk folder and the RFQ Register sheet.
2. Move the last 12 months of quotations and the current costing sheets into it, one sub-folder per customer.
3. Run the next five RFQs through the new routine and note the hours each one takes.

Pitfall: Do not rely on AI to read tolerances, material grades or finish notes from a drawing. A person checks every such value against the drawing itself.

2. Build a weekly owner dashboard from accounts, sales and stock data (start: days 31–60)

Why it fits: Your goal is to see orders, rejections and pending payments every week without chasing people.

Day to day: Every Saturday three people update their own tab in the Weekly Numbers sheet: the sales person (orders and quotations, from the RFQ Register), a production or quality supervisor (rejections by part and reason) and the accounts person (pending payments, from the Tally ageing report). Google Data Studio reads the sheet. The owner opens one page every Monday.

Tools: Google Data Studio (formerly Looker Studio), which is free and connects to Google Sheets, and Google Sheets inside your Google Workspace.

Effort: Medium. Building the charts is the small part. Getting three people to enter numbers every Saturday is the real work.

Benefit (estimate): Nobody compiles a weekly report by hand, the owner stops phoning around for numbers, and a problem shows up within a week instead of at month-end.

First steps:

1. The owner writes down the 5 to 10 numbers to review each week.
2. Build the Weekly Numbers sheet with one tab per source and one named person per tab.
3. Fill it by hand for three weeks before building any chart.

Pitfall: A dashboard built on late or inconsistent entries loses trust fast. Fix the Saturday habit first.

3. Send automatic payment reminders from your accounting data (start: days 1–30)

Why it fits: Payment follow-ups slip today, and pending payments are one of the three things the owner wants to see every week.

Day to day: Fully automatic reminders need a WhatsApp API platform or cloud accounting software, which is not worth buying yet for a B2B unit. Start with a weekly routine instead. Every Monday the accounts person exports the receivables ageing from Tally, gives Gemini the summary with customer codes instead of names, and gets reminder drafts graded by how overdue each bill is. Emails go from Gmail. Short nudges go from the WhatsApp Business app using saved quick replies, where the buyer's accounts contact is on WhatsApp. Bills overdue beyond a limit the owner sets go to the owner for a phone call.

Tools: TallyPrime, Google Workspace with Gemini, WhatsApp Business app.

Effort: Low.

Benefit (estimate): Every overdue invoice gets a reminder every week instead of when someone remembers, with less time spent writing each message.

First steps:

1. Check that every active customer ledger in Tally has the right accounts contact, email and mobile number.
2. Write three templates: gentle, firm and escalation.
3. Run the first Monday cycle and record each promised payment date in a sheet.

Pitfall: A reminder for a bill that is already paid, or with a wrong amount, hurts an OEM relationship. Reconcile receipts before sending. If the business is a Udyam-registered micro or small enterprise, the MSMED Act payment time limit (15 days, or up to 45 days under a written agreement) applies to its buyers; ask your CA whether to mention it in the firm template.

4. Give staff a business-grade AI assistant for drafting and summarising (start: days 1–30)

Why it fits: Some staff already use ChatGPT or Gemini individually, while the business handles financial records and customer contact details. Business tiers keep company data out of model training by default; many free consumer accounts do not.

Day to day: Office staff use Gemini from their company Google account for buyer emails (including English emails to export buyers), cover notes, meeting summaries, spreadsheet formulas and first drafts of SOPs. Personal AI accounts are not used for company work.

Tools: Google Workspace with Gemini first. ChatGPT Business, which needs at least two users, only if the heaviest users find Gemini falls short after a month.

Effort: Low.

Benefit (estimate): Noticeably less time on routine emails and letters for regular users. Ask each user to report hours saved once a month.

First steps:

1. The Workspace admin checks your edition and which Gemini features are switched on.
2. The owner adopts the one-page AI usage policy from this report.
3. Hold a one-hour session for office staff using the prompts in this report.

Pitfall: Without a clear rule, people keep using personal accounts on their phones. Name the approved accounts.

5. Extract data from purchase bills and expense receipts into accounts (start: days 31–60)

Why it fits: Manual data entry and reading invoices are both on your list, and accounts already run on Tally / TallyPrime.

Day to day: Supplier bills are scanned or photographed into one intake point, such as one email inbox or one Drive folder. The Docs by Ira AI plug-in in TallyPrime reads scanned invoices into draft vouchers. The accounts person checks GSTIN, invoice number, HSN, tax and total, then posts.

Tools: TallyPrime with the paid Docs by Ira AI plug-in, which needs a recent release and an active TSS subscription.

Effort: Medium.

Benefit (estimate): Most typing of purchase bills is removed. The accounts person's time shifts from entry to checking, with fewer keying errors found at month-end.

First steps:

1. Ask your Tally partner whether your release and TSS status support the plug-in.
2. Clean supplier ledgers and item masters. GST rates were rationalised from 22 September 2025, so check HSN and rate mappings at the same time.
3. Pilot with 100 bills and time it against manual entry.

Pitfall: Poor scans and handwritten challans read badly. Never post a draft voucher unchecked; a wrong GSTIN or tax figure affects input tax credit.

6. Plan reorders from sales history instead of guesswork (start: days 61–90)

Why it fits: You describe inventory and demand planning as guesswork, and procurement and inventory is a focus department.

Day to day: Once a month the purchase head exports item-wise sales and stock history from Tally into a Google Sheet. An AI assistant sorts fast and slow movers and suggests reorder levels and purchase quantities. The purchase head reviews them every week and adjusts for known customer schedules.

Tools: TallyPrime for the history; Google Workspace with Gemini or ChatGPT Business for the analysis.

Effort: Medium. It depends on whether item-wise stock is kept in Tally, which the questionnaire does not say. If stock is on paper, the first job is a stock sheet for the top items.

Benefit (estimate): Fewer stock-outs and less dead stock on the items covered, and quicker purchase decisions.

First steps:

1. Check that at least 12 months of item-wise data exists.
2. Record the supplier lead time for each of the highest-value items.
3. Set reorder levels for those items only, and review them weekly.

Pitfall: Where orders follow customer schedules or one-off jobs, past sales are a weak guide. Use this for regularly consumed raw material and consumables.

7. Create shop-floor SOPs and training in workers' languages (start: days 61–90)

Why it fits: Errors and rework are on your list, and supervisors are comfortable in Gujarati and Hindi.

Day to day: Each operation with frequent rework gets a one-page checklist in Gujarati and Hindi: setup, first-piece check, in-process checks, and what to do on a rejection. Gemini drafts and translates it from the rough or English original. The free Bhashini app can turn it into audio. Supervisors can ask Gemini Notebook questions that are answered only from the uploaded SOPs.

Tools: Google Workspace with Gemini, Bhashini, Gemini Notebook.

Effort: Medium. Writing down what experienced people know takes the time.

Benefit (estimate): New operators work unsupervised sooner, and fewer defects come from procedure errors.

First steps:

1. Use the rejection data from the dashboard to pick the three operations with the most rework.
2. Have the most experienced supervisor explain each one, and turn that into a draft checklist.
3. Translate it, have a supervisor validate it on the shop floor, then display it at the machine.

Pitfall: Machine translation gets shop-floor terms wrong. Keep gauge names, part numbers and tolerances in their original form, and have a fluent supervisor check every line.

Department-by-department plan

This section looks at the same roadmap from each department's desk: what hurts now, what changes, and the number to watch. Detailed steps are in the ranked opportunities and the 30/60/90-day plan.

Operations & delivery

Current pain: Errors and rework in routine work, with much shop-floor information on paper and phone. Your answers do not say where rejections are recorded, so the owner has no weekly view of them.

AI opportunity: Two things, in this order. First, a weekly rejection entry by supervisors in the Weekly Numbers sheet: part, operation, quantity produced, quantity rejected and a reason code. Column headings can be in Gujarati and Hindi. This is not AI, but the dashboard and every later improvement depend on it. Second, from days 61–90, Gujarati and Hindi checklists for the operations with the most rework, with audio versions for operators who prefer listening.

Tools: Google Sheets in Google Workspace with Gemini, Gemini Notebook (formerly NotebookLM), Bhashini.

First step: Agree five or six fixed reason codes with the supervisors, for example dimension out of tolerance, surface finish, material defect, setup error and handling damage. Record rejections against them for two weeks.

Metric to track: Rejected quantity as a share of quantity produced, by reason, each week. Later, the time a new operator needs to work unsupervised.

Not yet: Reading customer purchase orders and delivery challans straight into an ERP. That needs an ERP with clean item and customer masters and an implementation partner. Revisit it after the TranZact free-plan trial.

Sales & lead handling

Current pain: Quotations take two to three days because the inputs are scattered. Client follow-ups slip too, and nothing lists open quotations in one place.

AI opportunity: The RFQ Desk folder and the RFQ Register put every drawing, costing sheet and past quote in one place. Gemini Notebook finds similar past quotations, and Gemini drafts the quotation and cover email. For export buyers, who write in English, Gemini drafts clear English replies and clarification emails for the person handling exports to check. Once a week, filter the RFQ Register for quotations that have been in the Sent stage for more than seven days and follow up.

Tools: Google Workspace with Gemini and Gemini Notebook. TranZact on its free plan in days 61–90.

First step: Make one rule: an RFQ that arrives in a personal WhatsApp chat is saved to the RFQ Desk the same day by whoever received it.

Metric to track: Hours from RFQ received to quotation sent, the share sent within one working day, and quote-to-order conversion.

Not yet: A separate CRM. The RFQ Register does that job until it has been used consistently for three months.

Procurement & inventory

Current pain: Reorder quantities are guesswork, and purchase bills are typed in by hand. The questionnaire does not say whether stock is tracked in Tally or on paper.

AI opportunity: Reorder levels and suggested purchase quantities from 12 months of item-wise history, reviewed weekly by the purchase head. Draft purchase vouchers from scanned supplier bills also give a cleaner purchase history to plan from.

Tools: TallyPrime (for the history, and the Docs by Ira AI plug-in for bills), and Google Workspace with Gemini or ChatGPT Business for the analysis.

First step: List the highest-value raw material and consumable items and write the supplier lead time against each. If lead times are not written down anywhere today, this list is useful on its own.

Metric to track: Stock-out incidents per month and dead-stock value, both for the items covered.

Finance & accounts

Current pain: Payment follow-ups slip, bills are entered by hand, and reading supplier documents takes time.

AI opportunity: The Monday reminder routine from the Tally ageing export, with Gemini drafting and the accounts person sending. A pilot of Docs by Ira AI for purchase bills in days 31–60. Before paying for anything else, check two TallyPrime features that may already be available on your release: Connected Banking, which auto-reconciles bank statements, and Connected GST, which files GSTR-1 and GSTR-3B from inside Tally.

Tools: TallyPrime, Gmail with Gemini, WhatsApp Business app.

First step: Correct the accounts contact, email and mobile number in every active customer ledger. Then save three reminder templates as email drafts and WhatsApp quick replies.

Metric to track: Days sales outstanding and the share of invoices paid within terms, measured monthly. For the pilot, minutes of data entry per 100 bills and errors found at month-end.

Keep in mind: Bank statements and PAN or GST records stay inside Tally and company accounts. Only an ageing summary with customer codes goes into an AI assistant.

Owner / management reporting

Current pain: Reports are slow or missing, and the owner has to chase people for numbers.

AI opportunity: One Weekly Numbers sheet feeding one Google Data Studio page with orders, rejections and pending payments. Gemini can write a five-line summary of what changed since last week from the totals. The larger change is ownership: one named person per number, entered every Saturday and reviewed every Monday.

Tools: Google Data Studio (formerly Looker Studio) and Google Sheets in Google Workspace with Gemini. TranZact has an AI Data Agent that generates dashboards from factory data, but that only matters if an ERP is adopted after day 90.

First step: Write the 5 to 10 numbers on one page and put a name against each.

Metric to track: Whether the dashboard is reviewed every Monday, and the time spent preparing the weekly report.

Also the owner's job: Adopt the AI usage policy, and decide at day 90 what, if anything, to buy.

Recommended tools and indicative costs

Indicative prices from vendor pages on the date shown. Prices change, so confirm with the vendor before buying.

How to choose between these tools

Your budget band is moderate, but the first 90 days need very little of it. Of the eight tools in the cost table, you already have three (Google Workspace, TallyPrime and the WhatsApp Business app) and three more are free to start (Gemini Notebook, Google Data Studio and the Bhashini app). Keep the budget for what the first 90 days prove you need.

Gemini or ChatGPT Business? Start with Gemini. It sits inside the Gmail and Drive accounts staff already open every day, so there is no extra login to manage. What it can do depends on your edition: Starter includes Gemini in Gmail and the Gemini app, while Standard and Plus add Gemini in Docs and other apps, plus Gemini Notebook. If you are on Starter, compare two options for the few people who prepare quotations: upgrading their edition, or ChatGPT Business seats (minimum two users). Do not pay for both for the same person. Both list Gujarati and Hindi among their languages.

Gemini Notebook: Free with a Google account, with up to 50 sources per notebook. For company quotations and SOPs, use it through company Workspace accounts where your edition includes it. Do not move company files to personal Google accounts to get it.

Google Data Studio or a paid BI tool? Data Studio is free and reads Google Sheets, which is all the Weekly Numbers sheet needs. A paid BI tool adds nothing until entries are reliable.

Docs by Ira AI in TallyPrime: A paid plug-in sold in page packs. Start with a small pack for the pilot.

Languages: Gemini drafts and translates Gujarati and Hindi. The free Bhashini app adds translation and text-to-speech on a supervisor's phone. Sarvam AI APIs are strong Indian-language building blocks but need a developer, which does not suit medium tech comfort.

What not to buy yet:

• A paid TranZact plan. The free plan (100 enquiries a month, 3 users) is enough to test whether its enquiry-to-quotation flow suits you.
• A WhatsApp API platform such as AiSensy or Interakt. Weekly B2B reminders do not need one.
• A document-AI platform such as Nanonets, or a CRM.

Revisit all three at day 90, with numbers from the RFQ Register.

Your 30/60/90-day plan

The roles below are generic. Put a name against each one in week 1.

Days 1–30: quotations, reminders and safe AI use

Week 1

• Owner: Name four owners: RFQ Desk, costing sheet, Weekly Numbers sheet and AI policy. Adopt the AI usage policy from this report.
• Workspace admin: Check the Google Workspace edition and which Gemini features are on. Tell the owner whether the edition includes Gemini Notebook, because that decides the week 2 route. Confirm every office user has a company account.
• Person handling enquiries: Create the RFQ Desk folder and the RFQ Register sheet with its seven stages: Received, Costing, Draft ready, Approved, Sent, Won, Lost.

Week 2

• Person handling enquiries and costing owner: Move the last 12 months of quotations, current costing sheets and price lists into the RFQ Desk, one sub-folder per customer. Save each old Excel quotation as a PDF or Google Doc, named by customer code, part number and date.
• If the edition includes Gemini Notebook: Load the saved quotations into it from a company account. Plan for 50 sources per notebook, the free-tier limit, unless your edition shows a higher one. One combined file per customer keeps the count down.
• If it does not: Find similar quotes by hand. Search the RFQ Desk by part number and filter the RFQ Register by customer. Do not use a personal Google account to get round this.
• Accounts person: Correct the accounts contacts in Tally customer ledgers. Write three reminder templates. Run the first Monday ageing export and reminder cycle.

Week 3

• Person handling enquiries: Put every new RFQ through the quotation SOP in this report. Log received and sent times in the RFQ Register.
• Owner: Hold a one-hour session for office staff on the policy and the prompts.

Week 4

• Owner and costing owner: Review the RFQ Register. How many quotations went within one working day, and where did the others get stuck? If the by-hand route was used, decide now whether to upgrade the edition for the few people who prepare quotations or to add ChatGPT Business seats for them.
• Accounts person: List the customers who now need a call from the owner.

Measure: Share of quotations sent within one working day of the RFQ, from the RFQ Register. Reminder cycle run every Monday, yes or no. Number of office staff using company accounts for AI.

Days 31–60: weekly dashboard and bill extraction pilot

Week 5

• Owner: List the 5 to 10 numbers to review every week.
• Dashboard owner (an office person comfortable with Sheets): Build the Weekly Numbers sheet with three tabs: Orders, Rejections and Payments. One named person per tab.
• Supervisors: Start entering rejections with fixed reason codes.

Weeks 6–7

• The three data owners: Fill their tabs every Saturday.
• Dashboard owner: Build the first page in Google Data Studio (formerly Looker Studio).
• Owner: Review it every Monday.
• Accounts person: Ask the Tally partner about the TallyPrime release, TSS status and the Docs by Ira AI plug-in. Clean supplier ledgers and item masters.

Week 8

• Accounts person: Pilot 100 purchase bills through draft vouchers. Time it against manual entry and count the corrections.
• Owner: Decide whether to continue the plug-in. Remove dashboard charts nobody reads.

Measure: Dashboard reviewed every Monday without anyone compiling a report. Time spent preparing the weekly numbers. Minutes of data entry per 100 bills, and errors found on checking.

Days 61–90: reorder planning, shop-floor SOPs and an ERP trial

Weeks 9–10

• Purchase head: Export 12 months of item-wise history from Tally, if it exists. Record supplier lead times for the highest-value items. Analyse the export with Gemini or ChatGPT Business and set reorder levels for those items only.
• Production supervisor with one office person: Pick the three operations with the most rework from the Rejections tab. Draft checklists in Gujarati and Hindi.

Weeks 10–12

• Supervisors: Validate each checklist on the shop floor before it is displayed.
• Person handling enquiries: Run real RFQs through TranZact on its free plan, alongside the RFQ Desk, and note what it does better or worse.
• Purchase head: Review the reorder suggestions weekly.

Week 12

• Owner: Hold a 90-day review of quotation turnaround, rejection trend, days sales outstanding and stock-outs. Decide what to buy, if anything; any decision to pay for an ERP comes only now. Check the Udyam registration status and look at the two schemes in the schemes table.

Measure: Stock-out incidents per month on the items covered. Number of operations with a validated Gujarati and Hindi checklist. A written yes or no on a paid ERP, with reasons.

Step-by-step SOPs for your first two use cases

These two SOPs cover the first two opportunities in the plan. Copy them into a Google Doc, put names against the roles and adjust the details.

SOP 1: RFQ to quotation within one working day

Purpose: Send an accurate quotation within one working day of receiving an RFQ, with every input and output kept in the RFQ Desk.

Roles

• RFQ coordinator: the person handling enquiries. Logs, drafts and sends.
• Costing owner: the one person who owns the costing sheet. Sets every rate.
• Approver: the owner, or a person the owner authorises.

Inputs: The RFQ email or WhatsApp message, the drawing and its revision, quantity and required delivery date, the costing sheet template, past quotations in the RFQ Desk, and current raw material rates.

One-time preparation

• Your old quotes sit in Excel files and WhatsApp chats. Gemini Notebook takes PDFs and Google Docs as sources, so save each past quotation as a PDF or Google Doc first.
• Name every file the same way: customer code, part number, date.
• Plan for 50 sources per notebook, the free-tier limit, unless your edition shows a higher one. One combined file per customer, or one notebook per group of customers, keeps you within it.
• Use Gemini Notebook only from company accounts, and only if your Google Workspace edition includes it. The Workspace admin checks this in week 1. Never use a personal Google account for this.
• If your edition does not include it, skip the notebook and do step 3 by hand until the owner decides whether to upgrade the edition for the people who prepare quotations.

Steps

1. Log. On the day the RFQ arrives, save the message and drawing in the customer's sub-folder in the RFQ Desk. Add a row to the RFQ Register: date and time, customer, part number, quantity, required date. Stage: Received.
2. Check completeness. Confirm that material, quantity, drawing revision, tolerances, finish and delivery place are all stated. If anything is missing, write to the buyer the same day; Gemini can draft the clarification email.
3. Find similar work. With Gemini Notebook: ask for the closest past quotes for this or a similar part, then open each source file it cites and confirm it is the right part. Without it: search the RFQ Desk by part number and filter the RFQ Register by customer.
4. Cost. Stage: Costing. The costing owner fills the costing sheet: material, cycle time, operations, tooling, outside processes, packing and freight. AI does not set any rate.
5. Draft. Give Gemini the final figures and terms from the costing sheet and ask for the quotation in the company format plus a short cover email. Stage: Draft ready.
6. Approve. The approver runs the quality checks below. Stage: Approved.
7. Send. Send from the company email account. Save the PDF in the RFQ Desk. Record the date and time. Stage: Sent.
8. Follow up. Every Monday, filter the register for quotations that have been in the Sent stage for more than seven days. Follow up, then mark each one Won or Lost with a one-line reason.

Quality checks before approval

• Part number and drawing revision match the RFQ.
• Quantity, unit and rate match the costing sheet exactly.
• HSN code and GST rate are current.
• Lead time has been confirmed with production, not copied from an old quote.
• Payment terms and validity period are stated.
• For export quotations: delivery terms, currency and packing are stated, and the person handling exports has read the draft.

When the AI output is wrong

• Gemini Notebook cites the wrong part or an outdated quote: ignore it, search the customer folder by hand, and note the miss in an AI errors tab of the register.
• The draft contains a figure that is not in the costing sheet: delete the draft and ask again with the figures pasted in and an instruction to change nothing. Never ask AI to recalculate a rate.
• The wording commits to something the factory cannot do: the approver rewrites it by hand.
• The RFQ Desk owner reads the AI errors tab once a month and fixes the cause, which is often a badly named file or a missing source.

SOP 2: Weekly owner dashboard

Purpose: Give the owner one page every Monday showing orders, rejections and pending payments, without anyone compiling a report or being chased.

Roles

• Dashboard owner: one office person comfortable with Google Sheets.
• Three data owners: the sales person (Orders tab), a production or quality supervisor (Rejections tab) and the accounts person (Payments tab).
• Reviewer: the owner.

Inputs: The RFQ Register and order records, the supervisors' rejection entries, and the receivables ageing report from Tally.

Steps

1. Saturday, Orders tab. Enter orders received this week (customer, part, quantity, due date), quotations sent and quotations still open.
2. Saturday, Rejections tab. Enter quantity produced, quantity rejected and reason code, by part and operation.
3. Saturday, Payments tab. Paste the ageing summary from Tally: customer, amount outstanding, days overdue and promised date.
4. Saturday, completeness check. The dashboard owner confirms that all three tabs carry this week's date and messages anyone who has not filled theirs.
5. Refresh. Open the Google Data Studio page and check that its totals match the sheet.
6. Summarise (optional). Give Gemini this week's and last week's totals, without customer contact details, and ask for five lines on what changed.
7. Monday review. The owner reviews the page with the three data owners for 30 minutes. Each decision goes into an Actions tab with a name and a date.
8. Monthly tidy-up. Remove charts nobody uses. Add at most one new number.

Quality checks

• The Payments total equals the total outstanding in Tally on the same date.
• Rejection share is always calculated the same way: quantity rejected divided by quantity produced.
• Dates are in one format, and no week is left blank.
• Column headings are never renamed, because that breaks the charts.

When the AI output is wrong

• The summary describes a trend the numbers do not show: the sheet is the record. Delete the summary and review the numbers directly.
• A formula suggested by AI gives an odd result: test it by hand on five rows before using it.
• Never let an AI tool write into the source tabs. People enter the data; AI only reads totals and drafts text.

Ready-to-use prompts for your team

Paste these into Gemini from a company Google Workspace account, or into ChatGPT Business if you add it. Replace everything in [square brackets] with your own details.

Do not paste customer personal data. That means no contact names, mobile numbers, addresses, PAN or bank details. Use customer codes such as C-014 instead of names, and add the names back in your own document. Every output is a draft for a person to check.

Sales & quotations

1. Find similar past quotations. Use in Gemini Notebook, loaded with your past quotations, as soon as an RFQ is logged in the RFQ Register.

Using only the uploaded quotations, list every past quotation for part [part number or description], or for parts with similar material [material], size [key dimensions] and operations [operations].
Show a table: customer code, date, quantity, rate per piece, lead time, source file.
If nothing similar exists, say so. Do not estimate a rate.

2. Draft the quotation and cover email. Use after the costing owner has finalised the costing sheet, never before.

You are helping a sales engineer at a CNC machining unit in Rajkot that makes auto and pump parts.
Draft (a) a quotation and (b) a cover email of at most 120 words, using only the details below.
Do not change, round or calculate any figure. If a detail is missing, write [MISSING] in its place.
Customer code: [code]
Part number and drawing revision: [part no. / rev]
Material: [material]
Quantity: [qty]
Rate per piece: [rate from costing sheet]
Tooling or development charge: [if any]
Lead time: [weeks, confirmed with production]
Payment terms: [terms]
Quotation validity: [days]
Other notes: [packing, freight, inspection]

3. Ask an export buyer for missing details. Use when an RFQ from an export buyer is incomplete and the reply must be in clear English.

Write a polite, clear email in plain English to an overseas buyer.
Thank them for the RFQ for [part description] and ask for these missing details: [list, e.g. material grade, drawing revision, annual quantity, delivery terms].
Keep it under 100 words, number the questions, and say the quotation will follow within one working day of their reply.

Finance & accounts

4. Weekly payment reminders. Use every Monday, straight after the receivables ageing export from Tally.

Below is a receivables ageing summary with customer codes only.
For each row, draft a payment reminder email of at most 80 words and a WhatsApp message of at most 40 words.
Tone: gentle if overdue under [30] days, firm if [30] to [60] days, and a request for a firm payment date if over [60] days.
Mention the invoice number, invoice date and days overdue exactly as given. Do not add interest, penalties or legal language.
[paste rows: customer code, invoice no., invoice date, amount outstanding, days overdue]

5. Check a batch of purchase entries. Use before posting a batch of draft vouchers, and again at month-end.

Below are purchase entries: supplier code, invoice number, invoice date, taxable value, GST rate, tax amount, total.
List (1) duplicate invoice numbers for the same supplier, (2) rows where tax amount does not equal taxable value times GST rate, (3) rows where total does not equal taxable value plus tax, (4) dates outside [month and year].
Show only the problem rows, with the reason. Do not correct anything.
[paste rows]

Procurement & inventory

6. Suggest reorder levels. Use once a month on the item-wise export from Tally; the purchase head reviews every suggestion.

Below are [12] months of monthly consumption, current stock and supplier lead time in days for each item.
For each item: classify it as fast, medium or slow moving; suggest a reorder level and a reorder quantity; show the calculation.
Flag items where demand is too irregular for history to be a reliable guide.
Use only the data given. Present the result as a table.
[paste rows: item code, 12 monthly quantities, current stock, lead time]

Operations & shop floor

7. Turn a procedure into a Gujarati and Hindi checklist. Use when preparing shop-floor SOPs. A fluent supervisor validates the result before it is displayed.

Convert the procedure below into a checklist of at most [10] short steps for a CNC machine operator.
Write it first in simple Gujarati, then in simple Hindi, in everyday shop-floor language.
Keep machine names, gauge names, part numbers, numbers and tolerances exactly as written, in English.
Put safety steps first and mark them clearly. Do not add any step that is not in the original.
[paste procedure]

8. Summarise the month's rejections. Use before the monthly quality review, on rows copied from the Rejections tab.

Below are rejection entries: part, operation, machine, shift, quantity produced, quantity rejected, reason code.
Show the five part-and-operation combinations with the highest rejection share, and the three most common reason codes.
For each, list questions a supervisor should check on the shop floor. Do not guess causes.
[paste rows]

Owner / management

9. Five-line weekly summary. Use on Saturday, once all three tabs of the Weekly Numbers sheet are complete.

Below are this week's and last week's totals from a weekly numbers sheet: orders received, quotations sent, quotations sent within one working day, quantity produced, quantity rejected, total outstanding, outstanding over [60] days.
Write five short lines on what changed and what needs the owner's attention on Monday.
Use only these numbers. Do not explain causes you cannot see in the data.
[paste totals]

Three habits that improve every prompt: say who the reader is, paste the exact figures and tell the assistant not to change them, and ask it to write [MISSING] instead of guessing. If an answer looks wrong, do not argue with the assistant; check the source document and start a fresh request.

Data protection and risks (DPDP awareness)

Your answers say the business handles two kinds of sensitive data: financial records (bank statements, invoices, PAN and GST data) and customer names, phone numbers and addresses. A unit of 51 to 200 people also holds employee records. Under the DPDP Act 2023, any digital data about an identifiable person is personal data. That includes the contact persons at your buyers and suppliers, and your own employees. There is no turnover threshold, so a manufacturing unit of your size is covered.

Timeline: The DPDP Rules, 2025 were notified on 13 November 2025. The main obligations (notice, consent, security, breach intimation and data-principal rights) start on 13 May 2027. Treat that as the current date and watch for amendments.

What not to put into AI tools

• Bank statements, PAN, Aadhaar or salary details, in any assistant.
• Customer or supplier contact lists exported from Tally or WhatsApp.
• Anything at all into free personal ChatGPT or Gemini accounts. Use company accounts only.
• Customer drawings and buyer price lists. These are not personal data, but they are confidential. Check whether any customer agreement restricts sharing them before uploading them to any AI tool, including Gemini Notebook.

For reminders and reorder analysis, use customer and supplier codes instead of names. Mask before pasting, and test with sample data first.

Checks before choosing a vendor

Your business stays responsible for what its tools do with the data it shares. Before adding a tool:

• Read the terms. Is your data used for model training? Prefer business tiers that say it is not.
• Ask where the data is stored and who at the vendor can access it.
• Check that you can export and delete your data, and set a retention period.
• Put security terms in the contract, and do not let a vendor or freelancer keep data after the work ends.
• Add the tool to a simple register: which tool holds which data, and who has access.

Where a person must review AI output

• Every quotation: rates, tolerances, lead times and terms.
• Every draft voucher from a scanned bill, before posting.
• Every payment reminder: amount, invoice and tone.
• Every Gujarati or Hindi work instruction or safety step, by a fluent supervisor.
• Any reply to a tax or legal notice, together with your CA.

Two more habits

• Use accounts contacts only for the purpose they were given for. Do not reuse them for marketing without fresh consent.
• Write a one-page breach plan: who decides, who informs the affected people, and who informs the Data Protection Board and CERT-In. As general awareness, affected people are to be told without delay, the Board is to get a detailed report within 72 hours, and CERT-In's 2022 directions require certain cyber incidents to be reported within 6 hours.

Confirm what applies to your business with your CA or a qualified lawyer. This section is general information, not legal advice.

AI usage policy template for your team

Copy this into a Google Doc, fill in the [brackets], have the owner sign it, and share it with every office employee and supervisor. It is written for a unit of 51 to 200 people where office staff use AI tools daily and supervisors use them occasionally.

AI Usage Policy: Shreeji Precision Components

Version 1 · Effective from [date] · Policy owner: [name and role]

1. Purpose

The company uses AI tools to prepare quotations faster, draft emails and payment reminders, summarise weekly numbers and prepare shop-floor instructions in Gujarati and Hindi. AI prepares drafts. People remain responsible for everything that is sent, posted or displayed.

2. Who this applies to

All office staff, supervisors and anyone else who uses an AI tool for company work, on a company or a personal device.

3. Approved tools

• Gemini, used from a company Google Workspace account.
• Gemini Notebook (formerly NotebookLM), used from a company account, for past quotations and SOPs.
• The Docs by Ira AI plug-in in TallyPrime, for draft purchase vouchers.
• The Bhashini app, for translating and reading aloud work instructions that contain no confidential data.
• [ChatGPT Business, if the company buys it.]

Personal accounts on ChatGPT, Gemini, Meta AI or any other AI tool must not be used for company documents or data. A new tool may be used only after the policy owner approves it and adds it to the tool register.

4. Data rules

Never enter into any AI tool:

• Bank statements, PAN, Aadhaar, salary or other employee records.
• Customer or supplier contact lists.
• Passwords, login details or OTPs.

Enter only in approved tools, and only as much as the task needs:

• Costing figures and quotation terms.
• Receivables ageing summaries, using customer codes instead of names.
• Item-wise purchase and stock data.
• Customer drawings, and only where the customer's agreement does not restrict it. If unsure, ask the policy owner first.

5. Human review

AI never sets a rate, approves a payment, or decides anything about the hiring, pay or discipline of an employee.

6. Telling customers

Every quotation, email and reminder is sent by a named person from a company account. The company does not use an automated assistant to talk to customers. If that changes, for example by switching on automated replies in the WhatsApp Business app, customers will be told that they are dealing with an automated assistant and how to reach a person.

7. When something goes wrong

• A wrong AI draft caught before sending: correct it and note it in the AI errors tab of the RFQ Register.
• A wrong document already sent: tell the policy owner the same day. The sender issues a written correction to the customer.
• Data entered into a tool that is not approved, a lost device or a leaked file: tell the policy owner immediately. The policy owner follows the company's breach plan.

Reporting a mistake early is expected and will not be held against the person who reports it.

8. Ownership and review

The policy owner is [name and role]. The policy owner keeps the tool register, approves new tools and briefs every new joiner. This policy is reviewed every six months, with the next review on [date], and in any case before 13 May 2027, when the main obligations of the DPDP Act begin.

Signed: [owner] · Date: [date]

Before adopting the policy, have your CA or a qualified professional read the data rules.

Vendor shortlist and questions to ask

All options below come from the knowledge base behind this report. In each category, the first name is the one the plan starts with. Costs for the recommended tools are in the cost table.

AI assistant

• Google Workspace with Gemini: already in your accounts; the Gemini features depend on the edition.
• ChatGPT Business: a separate shared workspace with admin controls for two or more users; business data is not used for model training by default.
• Claude (Team plan): suited to long documents and spreadsheets; this report's data lists only Hindi among Indian languages for it, so test Gujarati before choosing it.

Document question-answering

• Gemini Notebook (formerly NotebookLM): answers only from the uploaded sources, with citations; the free tier allows up to 50 sources per notebook.
• Adobe Acrobat AI Assistant: answers questions about PDFs with source attribution; an add-on to Acrobat plans, so useful only if staff already work in Acrobat.

Dashboards

• Google Data Studio (formerly Looker Studio): free, and connects to Google Sheets.
• Zoho Analytics: paid plans add conversational questions (Ask Zia) and connectors for ERPs; consider it if an ERP is adopted.
• Microsoft Power BI: builds reports from Excel and databases; a better fit where a business already runs on Microsoft 365, and your answers list Google Workspace.

Purchase bill extraction

• TallyPrime with Docs by Ira AI: works inside the software you already use; a paid plug-in that needs an active TSS subscription.
• Nanonets: a separate document-AI platform with email intake and exports; more flexible, but more setup and one more system to maintain.

Manufacturing ERP (decide after day 90)

• TranZact: cloud ERP for SME manufacturers covering enquiries, quotations, purchase, inventory, BOM and production, with Tally integration and a free plan.
• It is the only manufacturing ERP in this report's knowledge base. Before paying, compare it with at least one other ERP that your Tally partner or other Rajkot units use.

WhatsApp

• WhatsApp Business app: what you use now; quick replies and labels are enough for weekly reminders.
• AiSensy and Interakt: Indian WhatsApp API platforms with broadcasts and a shared team inbox; worth it only if reminders must run automatically for a large number of customers.

Indian-language tools

• Bhashini: free government app for translation between Indian languages and text-to-speech.
• Sarvam AI APIs: speech and translation building blocks for Indian languages; they need a developer.

Questions to ask any vendor

1. Where is the data stored, and in which country?
2. Is the company's data used to train AI models, and where do the terms say so?
3. What are you doing to be ready for the DPDP Act obligations that start on 13 May 2027, and will you put security terms in the contract?
4. Does the product work in Gujarati and Hindi, both on screen and in the AI features?
5. What are the support hours, and is phone support available in India?
6. What is the minimum contract length, and can the plan be reduced or cancelled mid-term?
7. Can all data be exported to Excel or CSV at any time, and what happens to it after cancellation?
8. How much notice is given before a price change, and which features or usage limits cost extra?
9. Does it connect to Tally and Google Sheets directly, or through a partner at extra cost?
10. Who does the setup and training, and how long does it take for a unit of this size?

Relevant government schemes

Scheme details change often. Check the official page and your eligibility before applying.

Next steps this week

Seven actions for this week. None of them needs a purchase.

• ☐ Owner: Name the four owners: RFQ Desk, costing sheet, Weekly Numbers sheet and AI policy.
• ☐ Workspace admin: Check which Google Workspace edition the business is on and which Gemini features are switched on. Confirm every office user has a company account.
• ☐ Person handling enquiries: Create the RFQ Desk folder on Google Drive and the RFQ Register sheet with its seven stages: Received, Costing, Draft ready, Approved, Sent, Won, Lost.
• ☐ Person handling enquiries: Save every RFQ that arrives this week into the RFQ Desk on the day it arrives, including those that come on WhatsApp, and log the time received.
• ☐ Accounts person: Export the receivables ageing from Tally on Monday. Check that the ten largest overdue customers have a correct accounts contact, email and mobile number in their ledgers.
• ☐ Owner: Read the AI usage policy template, fill in the brackets, and tell staff that personal AI accounts are not to be used for company documents.
• ☐ Accounts person: Ask the Tally partner which TallyPrime release is installed and whether the TSS subscription is active.

At the end of the week, count two things: how many RFQs came in, and how many were logged in the register the same day. That count is the starting point for the 30/60/90-day plan.

Important information about this report

This report was generated automatically by software using AI. No human reviewed it before delivery. It is general information only, not professional, legal, tax or financial advice. Verify prices, vendor terms and legal requirements before acting.