AgileMorph Solutions

The AI Adoption Playbook

A 90 day plan and twelve prompts for the company you join, or the one you start. Built for the students of Badruka School of Management.

Session by Umang Dhandhania, Founder and CEO, AgileMorph Solutions. 4 August 2026.

The first 90 days

The most common mistake is walking in during week two with a grand AI strategy. You have no credibility yet and it gets ignored. Earn the mandate with one small, measured win.

Day 1 to 30

Observe. Propose nothing.

  • Map how work actually flows, not the org chart version. Sit with the people doing it.
  • Find the process everyone complains about. That complaint is your mandate later.
  • Note where data already exists and where it does not. This decides what is even possible.
  • Resist the urge to present an AI strategy in week two. You have no credibility yet.
Day 31 to 60

Pick one thing. Measure it first.

  • Choose a process that is high volume, rule based, low judgment, and low risk if it fails.
  • Measure the baseline BEFORE you change anything. Hours, error rate, turnaround time.
  • Run a small pilot with a defined end date and written kill criteria.
  • Keep the scope embarrassingly small. Cheap failure is the point.
Day 61 to 90

Show the number. Earn the mandate.

  • Present before and after in rupees and hours, not in features.
  • Name a business sponsor who benefits. A pilot with no owner dies quietly.
  • Write down what did not work. It buys more trust than the success does.
  • Only now propose scaling, and only to the next adjacent process.
Nobody gets budget for "AI transformation." Everybody gets budget for "I saved us 40 hours a month, here is the proof."

Start here

The CRAFT formula

Every good prompt has five parts. Most managers prompt badly and never find out.

Context: I run [FUNCTION] at a [INDUSTRY] company with [SIZE] people.
Role: Act as an operations consultant who has done this 50 times.
Action: Tell me which of our weekly tasks are worth automating first.
Format: A ranked table with reasoning, maximum 5 rows.
Tweak: (after reading) Now redo it assuming we have no budget for new tools.

Interrogate your own plan

The single highest value habit in this list. Use it before every decision that costs money.

Here is my plan:
[PASTE YOUR PLAN]

Act as the most sceptical person in the room. What is the single biggest
assumption I am making? What evidence would disprove it? What is the most
likely way this fails that I have not considered?

Find the opportunity

1. Audit your team's work

Run this in your first month. It turns a vague feeling into a ranked shortlist.

I work in [FUNCTION] at a [INDUSTRY] company with [SIZE] people.
Here are the 10 things my team does every week:
[LIST THEM]

For each one, rate: volume (high, medium, low), how rule based it is, how
much human judgment it needs, and the cost of an error.
Then rank the top 3 candidates for automation and explain your reasoning.
Flag any where automation would be a bad idea and say why.

2. Establish the baseline

The step almost everyone skips, and the reason most pilots cannot prove anything afterwards.

I want to automate [PROCESS]. Before I change anything I need a baseline.

Tell me exactly what to measure over the next 2 weeks, how to measure it
without disrupting the team or making them feel watched, and precisely
which numbers I will need later to prove the change worked.

3. Design a pilot that can fail cheaply

Small scope, fixed end date, written kill criteria. Decide how you will stop before you start.

I want to run a 6 week pilot automating [PROCESS] with [N] people.

Design it: what to include and what to deliberately exclude, success
criteria defined up front, what to measure weekly, and the kill criteria
that would tell me to stop. Keep it small enough that failure is cheap.

Build the business case

4. The ROI model

Nobody funds AI transformation. Everybody funds a proven number.

A team of [N] people spends [HOURS] per week each on [PROCESS].
Loaded cost is approximately [RATE] per hour.
An automation would cost [SETUP COST] to build and [MONTHLY] per month.

Calculate payback period and first year ROI. Then list 5 non obvious costs
or risks I am forgetting, and tell me what would have to be true for this
to fail.

5. The one page memo

Turn your pilot result into something a board or a partner will actually read.

Turn these notes into a one page memo for [BOARD / CEO / PARTNERS]:
[PASTE YOUR NOTES]

Structure: the problem in one line, what we did, the measured result, what
it cost, what we recommend next, and the honest risks.
Maximum 300 words. No jargon and no hype.

6. Advantage or just efficiency

Cost savings get competed away. Know which one you are actually buying.

Act as a strategy consultant. In [INDUSTRY] in India, where is AI creating
genuine competitive advantage, and where is it only creating cost savings
that competitors will match within a year?

Give me 3 examples of each, and tell me which category most companies in
this sector are wrongly investing in.

Choose and buy

7. Build, buy or partner

Most companies build plumbing they should have rented.

We are considering [CAPABILITY]. Argue both sides properly: the case for
buying an off the shelf tool, and the case for building it ourselves.

Consider our size ([SIZE]), whether this is core to how we compete,
switching costs, and who maintains it three years from now.
Then give me your recommendation and the single question that should
decide it.

8. Vendor due diligence

Ask these before you sign anything. Under the DPDP Act this is a legal obligation, not a preference.

I am evaluating [TOOL] for my company in India. Give me 12 questions to ask
the vendor, covering: where our data is stored, whether they train models
on our data, DPDP Act 2023 compliance, access controls, uptime and support,
exit and data portability, how pricing can change, and what happens if they
are acquired.

Tell me which answers should be outright dealbreakers.

Land it with people

9. Answer the fear before it is asked

Most rollouts fail on people, not technology. Their first question is whether they are being replaced.

I am introducing [TOOL OR AUTOMATION] to a team of [N] people who currently
do [WORK]. They are worried about being replaced.

Write a short, honest message I can actually say out loud. Do not make it
corporate or evasive. Then list the 5 objections I should expect and how to
answer each one truthfully, including the ones where the honest answer is
uncomfortable.

10. Write the usage policy

Write this before an incident, not after one.

Draft a one page AI usage policy for a [SIZE] person [INDUSTRY] company in
India.

Cover: which tools are approved, what data may never be entered, when a
human signature is required, who to ask when unsure, and the consequences
of breaching it. Plain English, no legalese, something people will read.

If you are starting your own

11. Map your whole operation

One founder can now do the work of several. The discipline is knowing where not to.

I am starting a [BUSINESS] in [CITY]. Map the full operational workflow from
customer enquiry to payment collection.

For each step tell me: can AI do this today, SHOULD it, and what is the
cheapest tool available. Flag every step where a human must stay in the
loop and explain exactly why.

12. Hire or automate

The question every founder faces around the first plateau.

My [BUSINESS] is at [STAGE OR REVENUE]. I am personally spending [HOURS] a
week on [TASKS].

Should my next move be hiring someone or automating this? Give me the
honest tradeoff, including what automation genuinely cannot do and what I
would lose by not having a person in that seat.

Five rules, no exceptions

1

Automate the work, never the accountability

Liability does not transfer to a vendor or a tool. Define which decisions require a human signature, in writing.

2

No clean data, no AI

Most failed projects die at the data stage, not the technology stage. If the data is a mess, that is the project.

3

Measure before you touch anything

Without a baseline you cannot prove the change worked, and an unproven win gets no budget for the next one.

4

Never paste confidential data into a public tool

Client financials, salary data, personal information. In India the DPDP Act 2023 makes this a legal matter.

5

Kill pilots fast and without ego

The cost of a dead pilot is small. The cost of a zombie project everyone privately knows has failed is enormous.