Phenomeny

FRACTIONAL AI LEADERSHIP

Your team is not short of people. It is buried under work that no longer needs one.

Chasing information. Re-entering it. Reporting on it. Answering the same question for the fortieth time this week. Take those four things off your team and you have a bigger company without hiring a single person. That is the entire job. Two fixed days a week, working directly with you.

95%

of enterprise AI pilots return nothing measurable. MIT NANDA, 2025.

42%

of companies abandon their AI programme entirely.

2

clients at a time, so yours is never one of forty.

Rs 3,00,000 / month

2 FIXED DAYS A WEEK · 3 MONTH MINIMUM · 2 CLIENT SLOTS

Call 9990377727

Direct line to Deepak. No form, no sales team, no discovery call with a junior.

WHAT IS ACTUALLY HAPPENING

Four problems. One answer.

Your senior people spend half their week finding things.

Your quotes and reports get assembled by hand, every time, from scratch.

Your best process knowledge lives in one person's head and one person's pen drive.

Your customers ask the same twenty questions and wait for a human to answer.

You have been sold these as four separate projects. Four vendors, four tools, four invoices.

They are not four problems. They are one missing layer.

An assistance layer sits underneath the business you already run. It learns how work actually gets done, holds what your team knows, and handles everything around the decision so your people only make the decision. Not a hack bolted on the side. Not a chatbot on the website. Infrastructure, the same way your accounting system is infrastructure.

WHAT BECOMES POSSIBLE

Weeks, where the industry takes years.

December 2024. A small Dubai engineering company, LEAP 71, hot-fired a working aerospike rocket engine. It is one of the hardest engine geometries on earth, mastered by only a handful of teams ever. Their AI model, Noyron, took it from specification to test-ready hardware in a matter of weeks. Aerospace measures that same journey in years.

Nobody gave them a bigger team. They pointed AI at the right problem and the timeline collapsed.

You do not need a rocket engine. You need the same collapse applied to the thing that actually holds your company back. The two day quote that goes out in ten minutes. The report nobody assembles by hand again. The six years of know-how that stays when the person leaves.

THREE COMPANIES, THREE FIXES

Same team. Different ceiling.

The lab that stopped losing its own research

A chemical R&D lab, five people, three months into a new compound. A competitor poached the lead researcher and two others followed. The data lived in logbooks and personal pen drives, and it walked out with them. No security policy stops a determined person, so we changed what gets captured instead of who gets trusted. A system that learned how the work was actually being done, so the next person picks up exactly where things stood, with the reasoning and the results intact.

The company's knowledge became the company's asset.

The manufacturer who did not need the ERP

A Swedish manufacturer, mid-rollout on a heavyweight ERP a board member had pushed through. The real problem was inventory sitting unmanaged in parallel, non technical staff, and almost no IT support. A large ERP does not fix a coordination problem, it formalises it. What worked was one place for documents, teams, projects and leads, that you talk to in plain language. Ask the question, get the report. Stop chasing three people who are each promising something different.

One source of truth, and nobody had to be retrained to use it.

The hospitality business that got its evenings back

Drowning in WhatsApp and email. Availability checks, price quotes, the same questions on repeat. A generic chatbot fails here, because somebody still has to physically confirm the real answer. So the AI took everything around the decision and left the decision with the human.

Far more enquiries handled, by the same people, with judgment still in human hands.

Not one of these was a headcount increase. Every one was capacity the company already owned, released.

HOW I WORK WITH IT

AI is an enabler. Knowing where it stops is what makes it profitable.

I put this early on purpose, because it matters more than any pitch line. AI cannot do everything, and it is not free to run.

Every model is a trade-off, and the trade-offs change monthly.

We spend real money on frontier models every month, which is how we know what each one is genuinely good for. You do not put an expensive reasoning model on a job a cheap, fast one does just as well. Getting that mix right is how the programme pays for itself. Getting it wrong is the quietest way to burn budget with nothing to show.

AI is only as good as what you feed it, and someone has to police that.

These systems learn patterns from whatever you give them. Good information in, and your team trusts the output enough to use it. Someone technical has to sit between your data and the model from day one, because quality problems do not show up in one place. They show up at scale.

The first build should be the boring one.

AI that quietly handles retrieval, follow ups, admin and reporting pays back faster than anything ambitious. Ship that, let the team feel the difference, then build the clever thing on a foundation people already trust.

THE PART NO VENDOR WILL SAY

Someone has to decide what you should not build.

You have already been pitched by four vendors. Each demos well. Each sounds certain. None of them will tell you that two of those four projects should not exist, because that is not their job and it is not in their interest.

MIT's 2025 NANDA study found 95% of enterprise generative AI pilots produced no measurable return, and 42% of companies abandoned their initiatives outright. Model quality was not the reason.

40%The tool did not fit how the work actually gets done
33%People did not trust the output
23%Nobody inside the company owned the decision

Three decision problems. Zero technology problems. And a decision problem cannot be outsourced to the vendor selling the solution.

I have no product on this page. Which means when I say do not build that, there is nothing hiding behind it.

WHAT MONTH ONE LOOKS LIKE

One page. What to build, in what order, and what to kill.

Two days a week, four weeks. I spend it learning how your people actually work, not how the org chart says they work. You end the month with a single document that names:

  • The two or three things worth building, and the sequence
  • The things you should not build, stated plainly enough that you can repeat the reason to the vendor who pitched it
  • Where your data actually sits today, and who can already see it
  • Your exposure under the DPDP Act, ahead of 13 November 2026
  • What you already own and are not using

Most of what I find is not a tool you need to buy. It is the distance between what your team does by hand and what nobody needs to do by hand again.

WHO THIS IS FOR

Owner-run companies between Rs 25 Cr and Rs 300 Cr.

RevenueRs 25 Cr to Rs 300 Cr
Headcount50 to 500
OwnershipPromoter or founder still runs the business
Who I report toThe CEO, MD or founder. Nobody else.
Current situationVendors circling, no reliable way to judge them

I work with the person who signs. Not a committee, not a steering group, not a project manager appointed to manage me.

COMMERCIAL TERMS

Two slots. Rs 3,00,000 a month.

FeeRs 3,00,000 per month
Term3 month minimum
Capacity2 clients. Waiting list once both are filled.
Exit30 days' notice after the minimum term
IPEverything built for you is yours. Phenomeny's own product line stays Phenomeny's.

Two slots is arithmetic, not scarcity marketing. Two fixed days each is already four days of my week.

BEFORE YOU CALL

Do not call if:

  • Your decision-maker is a CTO reporting to a board. You need a peer, not a reviewer.
  • The company is PE owned and AI strategy is set at fund level.
  • Revenue is under Rs 25 Cr. The fee is out of proportion to the decisions on the table.
  • You already know exactly what you want built. You need a vendor, not a strategist, and I will point you to a good one.

Saying this now saves us both a meeting.

WHY ME

Twelve years inside the systems that run a business. Not advising on them from outside.

Aditya Birla Group. Then Flipkart. Then nearly five years at Amazon in AI and Innovation, including Amazon Go's Just Walk Out.

E-commerce at that scale is not one function. Procurement, warehousing, logistics, order management, marketing, CRM, post-sales and recovery, all live, all touching each other. When one part breaks you see precisely how it breaks the rest. Most consultants have seen one slice. I have worked the whole chain, where mistakes are expensive and get noticed the same day.

On Amazon Go, zero learning curve was not a tagline, it was an engineering requirement. A store with no app to learn and no queue to stand in. That is the bar I hold anything built for your team to. If your people need training to use it, it is not finished.

At Flipkart I built Smart Assist, a single interface pulling fragmented supply chain data together for operations teams. The same idea people now call an AI assistant, more than a decade before the phrase existed.

Read the full background

Two slots. Rs 3,00,000 a month.

Call 9990377727. Tell me where the business is slowing down. I will tell you on the first call whether AI is the answer, and if it is not, I will say so.

Call 9990377727deepak@pddt.in

FAQ

Frequently asked questions