Dhaval Kapadia spent twenty years inside institutional capital, at Citi, at Apollo, and running his own advisory firm. He now builds the AI platforms rewriting that work, and helps investment firms get them past their own defenses.
Years inside institutional capital. Citi. Apollo. His own firm.
In institutional and retail capital he reported on, portfolio by portfolio.
Of enterprise AI pilots produce no measurable P&L impact.
AI platform built and shipped. Not advised on.
Citi · Apollo Global Management · REAM Partners · Blue Skies AI Studio · TiE LA Angels
Roughly 95% of enterprise AI pilots produce no measurable impact on the P&L. The cause is not model quality and it is not a lack of budget. MIT's researchers called it a learning gap: the tools were deployed, and the organization never changed.
In private capital, that gap has a very specific shape. And it has a name.
Every investment firm draws an invisible line around its most valuable information. The LP data. The deal files. The diligence. The models. The judgment of the partner who has seen this cycle before.
Inside that line, information is an asset. Outside it, information is a liability. The line exists for excellent reasons, and every good firm defends it.
It is also the single reason AI does not land in this industry.
Artificial intelligence is only useful inside the perimeter. That is where the work is. And nobody in the building is authorized to move the line, so the tools sit outside it, summarizing public filings and drafting emails, while the actual work of the firm goes on exactly as it did before.
The firms that win the next decade will not dissolve the perimeter. They will redraw it deliberately, with governance, so that intelligence can operate inside it.
That is the work.
“I spent seven years writing the investor reports that AI now writes in ninety seconds. I know exactly what it gets wrong.”
Seven dimensions that determine whether AI will land inside an investment firm. Not one of them is about which model you choose.
| Component | The question it answers | |
|---|---|---|
| C | Confidentiality Posture | What is actually permitted to touch a model, contractually and legally? Not what the policy says. What the LPA says. |
| A | Architecture of Work | Which workflows are genuinely AI-addressable, in what order, and which are theatre? |
| P | Permissions | Who is allowed to act on an AI-assisted output, and at what threshold? If nobody knows, nobody acts. |
| I | Instrumentation | Can anyone in the firm tell whether it worked? If not, it will be defunded within eleven months. |
| T | Talent and Behavior | Will the deal team change what it does on a Tuesday? This is the one that kills most programs. |
| A | Assurance | Model risk, audit trail, and the governance posture that procurement and the LPs will demand. |
| L | Leverage | Build, buy, or partner. Purchased and partnered solutions reach deployment roughly twice as often as internal builds. |
Ninety minutes with your leadership team. The failure data, the diagnosis, and your firm's own readiness scored live, in the room, by the people who run it. Most engagements begin here.
Enquire →A scored diagnostic that benchmarks your firm against every other firm we have scored. Seven dimensions. A heat map by function: diligence, IC, monitoring, LP reporting, underwriting. You will know exactly where you stand, and against whom.
Score your firm →Twelve weeks. Two streams. The Desk track rebuilds the workflows you already run. The Frontier track builds the capability you do not have and could not build inside the existing structure. Your team presents to leadership at the end, to secure resources. The output is a portfolio of funded work, not a training certificate.
Enquire →No stock presentations. Every engagement begins with research into your firm, your competitors, and the question your leadership is actually arguing about.
Every week, Dhaval sits down with the operators, allocators, and builders actually deploying AI inside financial institutions. Not vendors. Not futurists. The people who have to answer for it.
We are scoring the industry. 40 firms so far. The first benchmark of its kind, published quarterly.
Get the report →TiE Global Summit · TiE Investor Summit · TiECon West · Chapman University · Plug and Play