Focus on skills AI can't easily replicate: judgment, relationship-building, strategy and leading change. Then become the person who owns your firm's data and the AI workflows built on it. That turns a threat into a career advantage, because AI is only as good as its data, and whoever runs it decides what the team does with the freed time.
I get asked a version of this on a lot of calls, usually a little nervously. So I'll be honest about the starting point. A lot of platform and ops work is assembly. You collect portfolio updates, summarize them, format them for the partner meeting, and do it all again next month. You pull a list of candidates, write up notes, draft the newsletter. That kind of work is exactly what AI tools are getting good at.
But that doesn't mean the job goes away. It changes shape, and the people who shape the change keep the best parts of it. Here's how I'd make sure that's you.
Sort your work into three buckets
Start with an honest audit. Write down everything you did last week, in 15 to 30 minute chunks. Then tag each task with one of three labels:
- Automate: software can do this end to end with a quick review from you.
- Assist: AI can do a first draft or the research, but the output needs your judgment.
- Own: this depends on trust, taste or a relationship, and you should keep it.
Here's what that looks like for a typical platform week:
| Task | Bucket | Why |
|---|---|---|
| Collecting monthly KPIs from portfolio companies | Automate | Pull from what founders already send; ask only for gaps |
| Taking notes in founder calls | Automate | Capture is ambient now; you skim the summary |
| Drafting the portfolio newsletter | Assist | AI drafts well, but you pick what matters |
| Matching founders to experts for office hours | Assist | AI can shortlist; you know who gets along |
| Deciding which company needs help this week | Own | Judgment and context from many sources |
| Running a founder dinner | Own | The value is the room you build |
| Pitching a new program to the partners | Own | Influence and timing |
When platform leads I talk with do this audit, a big share of the week tends to land in the first two buckets. That's good news. It's time you're about to get back.
Own the data, not just the tools
This is the most direct career move available right now. In a lot of the funds I talk with, nobody clearly owns AI. Partners experiment on their own. Associates use chat assistants for research. None of it feeds a shared picture of the portfolio, so it stays siloed and dies when its champion leaves. Whoever steps into that gap becomes important fast.
But aim at the right gap. It's tempting to think a chat assistant is enough: hand it the documents and let it work. It's not magic. It's still math, and the hard part is the data underneath. Agents built on messy data produce slop, and if one number in four is wrong, nothing can run end to end. A context window is not a database. You can't paste four years of data on 40 companies into a chat and expect it to perform.
So use a crawl → walk → run rollout:
- Crawl (weeks 1 to 4). Use AI tools for your own work: meeting notes, first drafts, research. Keep a simple log of hours saved each week. Save the outputs where the team can find them, not only in your chat history.
- Walk (weeks 5 to 10). Get one data set right before you automate anything. Portfolio metrics are the usual candidate. Put them in one place, with every number linked to the email or deck it came from. Pull from what founders already send, and ask them only for what's missing.
- Run (week 11 onward). Build workflows on top of that data: the weekly portfolio pulse, LP update drafts, routing founder asks. Keep a human spot-checking the output, and have tools draft rather than send. Report results to the partners. At this stage you're the firm's operator for AI, whatever your title says.
On tools, buy the plumbing and build on top. I don't meet a lot of people who are happy with their in-house build a few months later, once they're the ones maintaining it. A homemade tracker can be quietly wrong for months before anyone notices.
The log from the crawl stage matters more than it looks. When you go to the partners in month three, "I saved six hours a week and here's what I did with them" is a far better pitch than "we should use more AI."
For the team-level view, see how AI changes the VC platform team and what to automate on a VC platform team. If you're picking tools, building vs buying AI at a VC firm covers that choice.
Build the skills AI can't easily replicate
Once the assembly work moves to software, what's left is the part that was always the real job. Four skills matter most, and each one shows up in platform work in a concrete way.
Judgment. Knowing which of 40 portfolio updates actually needs a partner's attention. Knowing that a founder's cheerful update hides a runway problem. Practice by writing a one-line "so what" on top of every summary you send.
Relationship-building. The founder who texts you before the board meeting. The buyer who takes your intros because the last three were good. No tool builds these for you, and I'm bullish that they stay human for a long time. More in how VC professionals maintain their networks.
Strategy. Deciding what the platform team should stop doing, not just what it should add. Practice by bringing one "stop" recommendation to every quarterly planning session.
Leading change. Getting a partnership to change how it works is hard, because partners have habits. In my experience the blockers are mostly psychological and behavioral, not technical. The person who can roll out a new workflow without a revolt is valuable, with or without AI.
Make the freed time visible
Saving time is only half of it. If nobody sees what you did with the hours, the easy conclusion is that the role needs fewer hours.
So redirect the time on purpose, and say so. Spend it on things partners already value: more founder one-on-ones, a sharper portfolio review, a talent program you never had room for. Then name the trade in your monthly update: "Metric collection now runs on its own, so I used that time to run four extra founder sessions on pricing."
This is much easier if the GP has given you explicit goals, because the freed time then maps to a number the partners already agreed matters. No goals yet? Ask for them.
That one sentence in the update changes how partners see your role. You go from the person who compiles things to the person who decides what's worth doing.
Three traps to avoid
The first trap is waiting. Some platform people keep their heads down and hope the tools pass them by. Then a partner reads an article, buys a tool, and hands it to the team with no input from the people who know the workflow. Now you're running a process someone else designed.
The second trap is becoming the demo person. You know the one: tries every new tool, posts screenshots in Slack, never ships a workflow anyone relies on. In my experience, a demo gets you 60 to 70% of the way. The last 30% takes months, and it's the part people actually rely on. I'll take one boring, reliable automation that saves the team a day each month over ten clever experiments.
The third trap is trusting output you can't trace. AI summaries can look exactly right and still be wrong (this happens more than people admit). If you can't show where a number in your partner summary came from, one bad figure will cost you more credibility than the tool ever saved.
The test for all three is simple. Each quarter, can you name one workflow that runs better because of you?
Will AI replace VC platform roles?
AI is unlikely to replace platform roles outright, but it is already taking over parts of them. Data collection, meeting notes, newsletter drafts and first-pass research are moving to software. The roles that remain will spend more time on relationships, judgment and programs, and less on assembling spreadsheets.
The risk is real for one kind of role: a job that's mostly assembly with little judgment on top. If your week is 80% collecting and formatting, that work will shrink, and the role with it, unless you move up.
The flip side is real too. A small team with good data covers far more ground than it used to. That's a strong argument for keeping, or even growing, a team that shows real outcomes.
Which VC operations skills are most valuable now?
The most valuable skills now are judgment about what matters, the ability to design and run a process end to end, clear writing, and comfort with data and AI tools. A platform or ops person who can define a workflow, automate the dull parts and explain the result to partners is hard to replace.
If you want a checklist for the next year, here's mine:
- Process design. Can you draw your firm's portfolio data flow on one page, from request to LP report?
- Data comfort. Can you define a metric clearly, such as "net revenue retention, trailing 12 months, from the company's own reporting," and spot when two companies report it differently?
- Writing. Can you turn a 20-page board deck into a five-line partner summary?
- Provenance. Can you show where every number in your last partner summary came from?
- Tool fluency. Can you set up and maintain an automated workflow without an engineer?
- Change management. Can you get five partners to adopt one new habit?
Pick your two weakest and work on them for a quarter. I'd bet on that over worrying about which job title survives.