SIRTIKA AI Blueprint

Where to start with AI — and in what order.

A paid diagnostic that tells a mid-market enterprise which department should adopt AI first, in what order, and what must be fixed before anything starts.

“This is not a report. It is a sequencing decision — what to build first, what to build second, and what not to build at all.”

Prepared by Naveenn Suri, SIRTIKA Creator · n.suri@sirtika.com · Sample built on a fictional client, Kalindi Pharmaceuticals.

The Moment

Meet Kalindi Pharmaceuticals

A ₹1,400 Cr formulations manufacturer — plants in Ahmedabad and Sikkim, regulated by CDSCO, Schedule M, and the US FDA at one site. Good company, real data, capable leadership. The board keeps asking about AI. And still, nobody can answer the only question that matters: where do we begin?

What is sitting on the desk

1A vendor has proposed a ₹3.2 Cr agentic quality platform — a tool demo dressed as a strategy.

2A 2024 forecasting pilot quietly died — nobody is quite sure why.

3Every deviation costs ₹2.8 lakh to investigate and close, and eight occur every month.

4270 Copilot seats and an OCR licence sit bought and unused — ₹42 lakh a year.

The Three Arguments

Every client has the same three conversations

“Our data isn't that bad — can't we just start a pilot?”

Below the data threshold, a model has nothing to learn from. The Data Gate protects you from the pilot that dies in month five.

“I expected Sales to be first. Why is it Finance?”

Because your own answers said so. The engine ranks on evidence, not on expectation — and the report addresses your pick by name.

“Why are you telling us what NOT to do?”

Because a company that knows what to ignore has been given something rarer than a recommendation.

The Answer

Then Kalindi came to SIRTIKA

Not a training course, not another tool demo. One diagnostic engagement — two to three weeks — that scores readiness, ranks every department, and sequences AI adoption into waves. It runs in three moves.

Diagnose

Six foundations

Score the organisation's real capacity to absorb AI — data, leadership, systems, people, governance, existing AI.

→

Rank

Ten departments

Score every function on value and feasibility, multiply to a Priority Index, and let the engine pick the top three.

→

Sequence

Three waves

Wave 1, Wave 2, Wave 3 — and a Do-Not-Start list, each call defended by the company's own evidence.

The Instrument

One instrument, many voices

198 questions in two parts. It is deliberately not answered by one person — nobody in a ₹1,400 Cr company can credibly speak to production data quality, invoice cycle times and service volumes at once. Five to eight function heads respond; the advisor reconciles.

Part	What it covers	Who answers	Questions
Part 1	Readiness Core — the six foundations (Modules M1–M7)	CEO, COO, CIO	78
Part 2	Department Assessment — the same twelve questions each	Each function head	12 ×

The Six Foundations — what gets scored

Every Part 1 question maps to one of these. They describe capacity to absorb AI, not appetite for it. Weighted to a single Readiness Score out of 100.

D25%

Data Foundation

Master data, historical depth, document reachability. The heaviest weight — and the one gate that can stop everything.

L20%

Leadership & Sponsorship

Whether an executive owns the outcome, and whether the last technology bet actually completed.

S15%

Systems & Integration

ERP, CRM, the reporting layer, and whether anyone has built against an API before.

P15%

People & Adoption

Grassroots AI usage, documentation discipline, and how the workforce met the last change.

G15%

Governance & Compliance

DPDP readiness, accountability for automated decisions, and audit obligations — weighted harder if regulated.

E10%

Existing AI Footprint

Licences owned, seats unused, pilots that stalled — nothing is recommended that the company already owns.

The three gates

Three rules run automatically once the answers are scored. They exist to stop a pilot that cannot succeed.

Data Gate

D < 4.0

No department enters Wave 1. The first ninety days go to data, not to AI.

Sponsorship Gate

L < 3.0

The roadmap stops at Wave 1. A twenty-four-month plan with no sponsor helps no one.

Governance Gate

G < 3.0 · regulated

Wave 1 is restricted to internal, non-customer-facing use until controls are fixed.

Four question types

SCOREDDescribe where you genuinely are today, rated against your revenue band.

METRIC+A specific number, percentage or rupee value you can defend.

CONTEXTSituation in your own words — not scored, but everything downstream depends on it.

GATEDetermines which sections and departments are issued next.

Inside the Readiness Core — open a module

Seven modules, each addressed to the person who can actually answer it. Click any module for a real sample question and the hint that guides a strong answer.

Part 2 — every department, the same twelve questions

That symmetry is what lets the analysis rank functions fairly. The twelve questions are really five measurements wearing different clothes — three decide whether it is worth doing, two decide whether it can be done.

Q1–2

Volume & repetition

How much of the same work repeats

Q3–4

Data availability

Can you produce the records next week

Q5–6

Cost of error & delay

What a mistake and the waiting cost

Q7–8

Tooling

A real system, or a spreadsheet nearby

Business impact

What changes if it works twice as well

Questions 10–12 are not scored — they supply the candidate use case, the person-hours, and the failure mode that appear in the report. Question 4 is decisive: a department that cannot produce its own data next week has its feasibility capped, whatever else it scores.

The Blueprint · Kalindi Pharmaceuticals Ltd

If they read nothing else

Every component of the 28-page blueprint, on one screen. Cleared to pilot AI in three named departments — chosen by the engine from the company's own answers.

Readiness Score

out of 100

Track

Pilot Ready

Track 1 · Band C

Completeness

91.4%

181 of 198 applicable

Confidence

High

all gates cleared

Executive Snapshot

Kalindi is cleared to pilot AI in three departments: IT & Quality, Finance & Accounts, and Supply Chain. The track is Pilot Ready, not Scale Ready, because Data Foundation sits at 6.2 and a strong composite does not override the data condition. Production — the one department the CEO named — does not enter Wave 1, because its batch records remain on paper until QA review. Net year-one value across Wave 1: ₹4.5 – 9.5 Cr (directional estimate).

Section 3

The six foundations, scored

The band matters more than the exact number — it describes how the company behaves today.

Readiness Score = (D×0.25 + L×0.20 + S×0.15 + P×0.15 + G×0.15 + E×0.10) × 10 = 63. Data carries the most weight because a model with nothing to learn from cannot be rescued by good leadership or good systems.

Section 4

Gate rulings — including the ones that did not fire

Gate	Threshold	This company	Ruling
Data Gate	D < 4.0	D = 6.2	Did not fire
Sponsorship Gate	L < 3.0	L = 6.8	Did not fire
Governance Gate	G < 3.0 · regulated	G = 6.8 · regulated	Did not fire

The Governance Gate was evaluated because Kalindi is a regulated manufacturer. Governance scoring 6.8 is what permits a customer-adjacent use case to be considered at all — it does not remove the GxP constraints in Section 11.

Section 5

The department heat map

Ten departments on two axes — worth doing, and buildable. Priority Index = Value × Feasibility. The engine selects the top three, not the client and not the advisor.

Department

Value

Feasibility

Index

Quadrant

Do First Opportunistic Build Toward Do Not Start

Strategic priority from L8, answered by the CEO: Improve compliance (1st) · Reduce risk (2nd) · Improve decision quality (3rd). The CEO ranked compliance first; the engine ranked Quality first. Those two facts are the same fact.

Section 6

The sequencing decision

Why these three

IT & Quality (63.3), Finance (60.9) and Supply Chain (48.1) combine high value with feasibility above 5.0, and all three answered Question 4 affirmatively.

Why not Inventory

It qualified (PI 42.7, Do First) but Wave 1 admits three and it ranked fourth. It sits at the head of Wave 2 — queued, not blocked.

Why not Production

Third-highest value at 6.7, but feasibility is capped at 4.0 by rule: batch records stay on paper until QA review. It cannot host a model that learns from history it reconstructs.

Kalindi should also decline the ₹3.2 Cr agentic quality platform proposed by a vendor. It solves the Quality problem in this blueprint at roughly six times the cost, and it assumes searchable deviation data Kalindi does not yet have.

Section 7

Wave 1 — the first ninety days

For each department: the use case, the technology category, the named owner, the value, and the failure mode. Click any card to open it.

Section 9

Foundation fixes

The work that must happen regardless of which department goes first. The last column is the reason this section exists.

Fix	Foundation	Effort	Owner	Unblocks
Decide the e-signature position for machine-drafted records	Governance	1 pm	Head of Quality	IT & Quality — the entire Wave 1 use case
Index the deviation and batch-record repository by content	Data	3–6 pm	IT Head	IT & Quality — retrieval quality
Replace email despatch confirmations with a structured feed	Data & Systems	2–4 pm	Head of Supply Chain	Supply Chain — this is why the 2024 pilot failed
Activate 270 dormant Copilot seats and the unused OCR licence	Leadership & Existing AI	1 pm	MD	₹42 lakh a year. The OCR licence also underwrites Wave 1.

Section 10

Investment and return

As ranges, every figure a directional estimate. Cycle-time recovery dominates — the money freed when work that took days takes hours.

₹6.45 – 10.46 CrGross annual value

−

₹0.92 – 1.91 CrFoundation, validation & build cost

₹4.5 – 9.5 CrNet year-one value

Payback on midpoints falls at roughly two to three months. Validation lead time does not compress: plan six to nine months to first benefit in Quality, four to six in Finance. Loaded hourly cost ₹824, from C6 ÷ 176 productive hours. Every figure is a range and a directional estimate.

Component	Low ₹ Cr	High ₹ Cr	Source
B1 — Manual effort recovery	1.02	1.54	Client data
B2 — Error and rework reduction	1.83	3.02	Client data
B3 — Cycle-time value	3.60	5.90	Client data
Gross annual value	6.45	10.46
B4 — Foundation, validation & build cost	(0.92)	(1.91)	Benchmark-derived
Net year-one value	4.54	9.54

HighConfidence

All three automatic gates cleared, and the Data Foundation (6.2) sits comfortably above the 5.0 pilot threshold. The company can fund Wave 1 now, not conditionally — provided batch records are digitised in parallel with the IT & Quality pilot.

Section 11

Guardrails

GxP & 21 CFR Part 11

Any system that supports a batch release sits inside the validated boundary. Wave 1 Quality is drafting only: the system drafts, the investigator concludes, the QA head signs. No automated entry into the quality record.

Data integrity — ALCOA+

A drafted deviation report is not contemporaneous and not original. It must be captured as an input to the investigation, with the investigator as author and the tool disclosed — or a 483 observation follows a productivity gain.

DPDP

A DPO is named and consent flows are mapped. No Wave 1 use case touches personal data. The HR document use case in Wave 3 does, and waits on the retention position for CVs and appraisals.

Shadow AI

Thirty-five percent of office staff use public AI tools; the policy is silent on batch and patient-adjacent data. In a US-FDA-inspected company that silence is the exposure, not the usage.

Section 12

The next thirty days

Decide whether a machine-drafted deviation record sits inside the validated boundary

Head of Quality + QA

14 days

Activate the unused OCR licence and the 270 dormant Copilot seats

21 days

Replace email despatch confirmations with a structured feed

Head of Supply Chain

30 days

Extend the AI usage policy to batch and patient-adjacent data

DPO

30 days

Decline or defer the ₹3.2 Cr agentic quality proposal

30 days

Appendix A

How the Priority Index was produced

Worked through on IT & Quality, the highest-ranked department. Every other department was scored the same way, from its own answers.

Step 1 — twelve answers become five measurements

8.0Volume

7.0Data avail.

8.0Cost of error

6.0Tooling

8.0Business impact

Step 2 — is it worth doing? (Value)

(8.0 × 0.30) + (8.0 × 0.30) + (8.0 × 0.40) = 8.00, then × 1.20 compliance priority = 9.60

Step 3 — could it be built? (Feasibility)

(7.0 × 0.40) + (6.0 × 0.25) + (6.2 × 0.20) + (7.0 × 0.15) = 6.59

Step 4 — the Priority Index

9.60 × 6.59 = 63.3 — multiplied, not added. A valuable thing you cannot build is worth nothing at all.

Section 13

Methodology and confidence

Questions answered: 181 of 198 applicable. All ten departments in scope. Seventeen unanswered, twelve of them in Production.
Completeness 91.4% · Confidence High. Calibration band C (₹1,000 Cr+), a −1 anchor shift applied — Kalindi is scored against enterprises, the correct bar at ₹1,400 Cr.
Inverted question set: D9, P6, G7, E5. Penalty: −0.5 to Leadership for ₹42 lakh of unused annual licence spend disclosed at E5.
Every ₹ figure is a range and a directional estimate. Client-data figures come from Kalindi's answers; benchmark-derived figures from published ranges applied conservatively. No study cited that has not been verified.

This is a sample. Yours is built from your own answers.

The live diagnostic reconciles five to eight of your function heads, applies the full calibration, and delivers a 28-page blueprint, a ranked Excel Opportunity Register, and a 60-minute debrief.

Book the debrief. Walk the sequence. Defend the Do-Not-Start.

Book the debrief

Naveenn Suri

SIRTIKA Creator

Book the debrief ↗ www.sirtika.com