The Partner's 90-Minute AI Briefing

1. Partners Need Decisions, Not a Tutorial

Most AI briefings aimed at leadership are tutorials in disguise. How the models work, what a token is, a tour of the latest features. Partners do not need any of that, and sitting through it is the fastest way to conclude AI is someone else's job.

What a partner needs is the short list of decisions only leadership can make, the information required to make them, and the freedom to stop thinking about AI the rest of the quarter. The mechanics belong to whoever owns the work. The direction, the money, and the message to LPs belong to the partners.

So this is built as a decision brief, not a class. By the end you should be able to say what the firm will start with, who owns it, what you will tell LPs, and what you will spend. Four answers. Ninety minutes.

2. What Is Actually Real Right Now

Cut the hype and the noise in both directions. AI is neither about to run your fund nor a passing fad. For an investment firm, a specific and narrow set of uses is genuinely real today, and the rest is still maturing.

Real now. Reading long deal documents and pulling out what matters. Drafting first-pass memos, IR letters, and DDQ responses in the firm's own voice. Answering questions across your own knowledge base. Automating recurring packs, the monthly portfolio update, the board summary, the standard report. These work, in production, at firms like yours.

Real, with a person on top. Anything where a number drives a decision. The model assembles and drafts, fast and well, but it is confidently wrong often enough that the output is a starting point a person verifies, never the final word. That is a feature of how you use it, not a reason to wait.

What is not yet real is the fantasy version: an autonomous system that sources, diligences, and decides on its own. Treat anyone selling that with the skepticism you would bring to any deal. The honest frontier is wide enough to matter without it.

3. What It Costs

The honest range is wide because it tracks ambition, not the technology. Here is what each level actually runs.

Getting going. Per-seat licenses for the team plus a focused training. Low thousands of dollars a month in software, plus a one-time training engagement. This buys most of the available value, which is the part vendors selling six-figure platforms do not lead with.

Running it deliberately. The above, plus someone accountable for adoption. If that is an outside partner rather than an internal hire, an AI Operating Partner retainer starts around ten thousand dollars a month. This is what turns licenses into changed behavior.

Building something custom. Automated workflows and connected systems are a project, not a subscription, and they start in the low-to-mid six figures depending on scope. Worth it only after simpler tools have proven the value, never as the first move.

The pattern to remember: cheap to start, and most of the value is reachable cheaply. The expensive end is real, but it is earned later, against the loaded cost of the hours it removes. The full breakdown by stage is in how much a fund should spend on AI.

4. The Decisions Only Partners Can Make

Almost everything about AI can be delegated. Four things cannot, because they are about the firm's direction, money, and reputation, which is the partners' job by definition.

Make these four and you have done the partner's part. Leave any one of them open and the initiative drifts, because the team will not pick the firm's direction, name its own mandate, write the LP line, or approve its own budget. Those are yours.

5. What to Approve This Quarter

A sensible first quarter is small, cheap, and concrete. Resist the pull toward a grand program.

Approve three things. Licenses for one team to work with a capable tool on real material. A focused training so they use it well rather than poke at it. And one named owner, given real time and your visible backing to win one workflow. That is the entire first move, and it is enough.

Pick a workflow that costs real hours, produces output a person can check, and uses data you are comfortable putting through the tool. Deal-document reading and first-draft memos usually qualify. The point is one clean win the firm can see, not a launch event. Where exactly to aim is the subject of where to start with AI.

Keep the first quarter cheap enough that approving it is easy and the only real question is whether it worked. You are buying a signal, not committing to a platform.

6. What to Say No To

Saying no is half the job, and the cheaper half. A few things deserve a clear no, for now.

No to the firm-wide rollout on day one. Everyone, everything, all at once gives no one a place to start, and it is the most reliable way to spend money and produce nothing. Win one workflow first.

No to the six-figure custom build before a single pilot has paid off. The expensive end is real and it comes later, after simpler tools have proven the value. Buying it first is how budgets die.

No to the tool with no owner. A license with no accountable person is a subscription you will cancel in a year, having learned nothing. If you cannot name the owner, do not approve the spend.

None of these is a permanent no. They are "not yet, and not like this." Sequencing is most of what separates the firms that get value from the ones that get invoices.

7. The Risk Partners Should Actually Worry About

The risk that gets the headlines is the wrong one to lose sleep over. Partners worry about the model going rogue or replacing the judgment the firm is built on. Neither is the live threat.

The real risk is mundane and far more common: the firm spends real money on tools nobody adopts. An MIT study in 2025 found about 95 percent of enterprise AI pilots delivered no measurable return, and the tools were rarely the reason. The money goes out, a few people try it, the energy fades, and a year later there is nothing to show. That is the failure mode that actually happens at firms, and it is entirely preventable.

The way you prevent it is the way you prevent any initiative from fading: a named owner, a narrow first win, and proof on real work before you scale. The detail of why rollouts die, and how to land in the 5 percent, is in the AI strategy and roadmap.

The data question is real too, and it is a governance decision, not a reason to stall. Set the rules for what goes through which tool, and the firm gets the upside without the incident. That is a line item in the briefing, not a project that blocks it.

8. How to Tell If It Is Working

Partners do not need a dashboard. They need one test.

If you turned the tool off tomorrow, would anyone be upset. If yes, it was adopted and it is working. If no, it was installed and the money is sitting idle, no matter how many people logged in once. Logins, messages, and seats flatter the report and tell you nothing about whether the work changed.

Ask that question one workflow at a time, a quarter in, and you will know the truth long before the year-end review. The fuller version of the measurement is in the strategy guide, but the off-switch test is the one a partner can keep in their head.

9. The 90-Minute Agenda

Here is the meeting, end to end. Ninety minutes, four decisions, no tutorial.

First fifteen, what is real. The narrow set of uses that genuinely work for the firm now, and the autonomous fantasy you are setting aside. Next fifteen, what it costs. The honest range by level, and the point that most of the value is cheap. Next thirty, the four decisions. Where to start, who owns it, what to tell LPs, what to budget, argued and settled in the room.

Next fifteen, what to approve and what to refuse. The small first quarter, and the clear list of no-for-now. Last fifteen, how you will know. The off-switch test and who reports it back, so the next review is about evidence, not opinion.

That is a real decision made in an hour and a half, by the only people who can make it. Everything after is execution, owned by someone who is not in this meeting after today. This is precisely the shape of a 90-minute Executive Briefing, run for your partners with your firm's context in the room.

10. Where to Start

Put the ninety minutes on the calendar. Use the agenda above, and do not leave the room until the four decisions have answers: where to start, who owns it, what to tell LPs, what to budget. A decision deferred is the initiative deferred.

Then approve the small first quarter, say no to the rest for now, and ask the off-switch question at the next partner meeting. That is the whole of the partner's job here, and it is enough to put the firm on the right side of the odds.

If you want this run for your partners rather than read off a page, that is exactly what a 90-minute Executive Briefing is: the honest picture, the four decisions, and your firm's context in the room, the entry point to everything that follows. When you are ready to turn the decisions into a sequenced plan, an AI Readiness Sprint produces the baseline, the first workflow, and the roadmap, and the line you give LPs is covered in what to tell your LPs about AI.