Our focus now shifts to structuring an FP&A team in a world of automated AI agents. Structuring a team that will build better outcomes, faster.
The proudest result I have had is when I rebuilt an entire finance and strategy team across a 10,000+ person company, to a level of success that I had nothing to do. It may be my proudest professional accomplishment. The team was killing it. Each of our eight countries had its own CFO who ran key decisions and drove analyses. My key deputy at the time managed those teams and a small global consulting team ran itself. Everyone was quite happy, so I left to find the next challenge. And their progress continued:
My deputy became the global CFO within the next two years
Nearly all of the team are still there after five years
The company’s achieved growth of 30%+ over each of those years
When did I know it was time to go? I rarely needed to be consulted for major decisions because the team owned them and was coming up with great ideas. The rolling budget process was working like clockwork. Models were providing direct guidance that led to better results. Emergencies were handled quickly and accurately without the need for me to jump in. I could take a month-long vacation without a fire.
It took about seven years of work by building several different teams at the company to achieve the success. But when it worked, it worked. Now, this process will be supercharged with the introduction of agents.
The new reality of an FP&A team: Less time spent on data reconciliation, report building, and scenario planning. Fewer people, who are more comfortable building automated tools, reviewing conclusions, and creatively forming conclusions.
To get to it, you need to re-structure your team (if you haven’t already) around the new reality.
Type of Work | AI Agent Work | The New Analyst Role |
Data Consolidation | Spot mismatches across ERP, CRM, billing. Push alerts with exact fix‑paths. Every system becomes the source of truth. | Ensure fixes get made by each team. Train the team on how not to make errors in the first place. |
Reporting | Draft decks and dashboards from a short prompt. Write first‑pass commentary for each audience. | Fact‑check, tighten the story, keep one narrative across the org. |
Planning | Roll the forecast forward using pipeline, usage, and spend. | Decide where to lean in or cut back. Align with leaders on priorities that software can’t guess. |
Scenarios | Create useful scenarios across any decisions quickly. Cascade the decisions and auto-update all numbers that add up to the scenario. | Review scenarios and decide with the relevant team what is the most realistic in terms of team members, actions, and next steps. |
Agents / Business Partners | Give an automated update each day on where the company is heading. Track across hundreds of reports instantly to see outliers. | Set up the agents and review what they flag. |
System Setups | Quickly format different types of data into a single format for any data uploads. Provide quick ‘how to’ to enable anyone to use the current tools quickly. | Create the right systems, teach other teams how to use AI for questions and reporting. |
All in all, AI agents will allow you to have fewer, stronger analysts focused on improving outcomes.
The Bullets
|
Let's get to work:
1. Discover your gaps
Your team should be reviewing and building automations that further the company. The first thing to do is to understand where you are now and what you want to improve. We have provided a sample below. To do this:
List out the tasks: To do this, create a chart of all the tasks (repetitive or not) for your team. Separate them into what you can do, what AI agents should do, and what an analyst will do.
Roughly break out the time and quality of each. How much time is your team spending on X (ex. AR rec; Monthly report spreadsheet; Board Material). What is the actual result of these tasks? Are they improving outcomes?
Decide on the final vision.
What will be automated? It may not be now. But anything repetitive should be in this bucket. For ideas.
Where will you spend the most on strategy? AI will help give you the options, but not make the decisions. Don’t forget the importance of business partner work (though, one business partner should be able to support many teams).
What technical work you can outsource? If you need expertise on certain things, like tax or legal treatment. Outsource. This allows you to hire generalists over specific experts.
Create a roadmap of how you will get there. There are only a couple of things you can improve at a time by either making it faster, making it more accurate, making it more persuasive. For help prioritizing, this matrix still holds.
Every team is different, but the idea is that you want to identify gaps, and then make them projects for your team to accomplish. Example below.
Tip: Don’t try to do every AI automation at once. Prioritize and start with a few small wins. Teams get better through success.
Type of Work | Specific Task | Time (Old → New) & Owner/How | Company Outcome (Current → Desired) |
Data Consolidation | Daily Performance Overview | 2 days → 1 hr (Import into FP&A platform, create daily summaries) | People wait until month-end → People always know current priorities |
Data Consolidation | Data Accuracy – Source Systems | 1 day/week → 10 min/day (AI checks errors daily) | Siloed systems → Always consolidated data |
Reporting | Monthly Reporting Sheet | 3 days → 30 min (NLG + templates) | Delayed insight → Finance closes fast, execs react same week |
Planning | Rolling Budget (Monthly) | 5 days → 3 hrs (Auto-roll + driver AI) | Static plan → Live budget that adjusts with reality |
Agents/Biz Partners | Sales Reviews | 6 hrs prep → 20 min (AI highlight pack) | Hunch-based coaching → Data-backed deals; higher win rates |
Scenarios | New Pricing Scenarios | 1 day → 15 min (Parametric bot) | Single guess → Multi-option ROI; pricing launches 3× faster |
Agents/Biz Partners | One-off Requests | 1 day turnaround → Instant (Chatbot answers) | Finance bottleneck → Self-serve decisions |
Scenarios | Cash Runway Test | 6 hrs → 20 min (Monte Carlo AI) | Rough cushion estimate → Probabilistic runway; timed funding |
Reporting | KPI Dashboards | 1 day refresh → Real-time (via BI API) | Outdated KPIs → Always current; board self-checks progress |
2. Design around these projects
If you haven’t already, you need to make sure that your team management is built around projects as opposed to general ‘areas of responsibility.’ This enables your team to implement automations and then move on. Done well, your entire team will be supercharged in a quarter. The ability to work with data and drive people to action is more important than in, say, Sales Support.
Plug the team around projects. Perhaps you have one person who is a ‘business partner’ and another analyst that tweaks and sets up the agents for one team. Make sure everyone is working on a project that will automate rote work.
Keep it flat. With modern FP&A tools, there’s no reason to have “senior managers” managing two people, like a ‘head of Revops.’ Instead, have the entire FP&A team report to you directly. This lets you build your team’s skills and stay in the loop on critical decisions.
Reset role expectations. Today’s FP&A analyst should be able to leverage AI to cover several departments. Time spent on data reconciliation, summarizing information, and brainstorming best practices should drop dramatically. Where you used to assume one business partner per department, now expect one business partner to cover at least three or four.
Ensure business partners report into you. Some teams may already have embedded business partners (sales enablement, marketing analysts, executive assistants). These should report into you. This gives teams a consistent experience, allows you to share automation learnings, and prevents the dreaded ‘analyst spread.’ And don’t hesitate to be one yourself (to Sales and CEO) so you stay close to the action.
You're building the team for the future. Imagine each of your team members as independent mini strategic CFOs.
Tip: Don’t overhire. Don’t overhire. At most startups the data work is relatively easy. Estimates show that FP&A teams are already 30% over-hired…
3. Make the right type of hire
You need fewer people who are truly comfortable using agents and who can provide consistent, reliable advice. You’re looking for the creative number guru who lets the bots do the grunt work, which they turn into insight and action. That means analysts who are:
Flexible and creative. Hire people who are creative, driven, and data oriented. Then, as you grow into a public company they will deepen in one area (pricing, SQL, revenue ops) as needed.
Somebody people trust. Numbers mean nothing if no one believes the messenger. Hire people who speak plainly and keep promises.
Curious. You want people who ask “why” until it hurts, try new tools on a Friday night for fun, and kill manual steps without permission.
Tech fluent. They should write clean prompts, audit the agent’s math, and patch errors fast. And be able to pick up new systems quickly using AI as a teacher.
Enjoy teaching. They document fixes in plain English and make sure the next person can run analyses with them.
Low‑ego, high‑ownership. They share credit, take blame, and always leave the process better than they found it.
The famous saying is that the best generals are ‘smart but lazy.’ You want people that are driven to automate everything as much as possible.
Tip: If you have fewer people, you can afford to pay each one more.
4. Getting your current team onboard
You’re usually not building from scratch. You already have a team. Speed and care matter. People might already be a bit nervous. This is your chance to set the tone, fix the narrative, and chart the path forward. Here’s how to approach it.
Excite the ship. Paint a vision of the future where everyone can contribute more to strategic insights and gain technical expertise.
People first, AI tools. Duolingo’s CEO had a good starting post. You want the team to work and learn together in these changing times.
Get others’ feedback. Talk to every stakeholder who uses finance outputs. Ask what slows them down and what they wish they had. Write it up in plain English and share it back that same week.
Give a one-month PR. Give a one-page feedback on wins, gaps, and next priorities in terms of the new projects. Keep it honest and number-heavy and quick.
Make decisions in 3 months. Lock the tool stack. Set the rolling forecast cadence. Anything still manual? Schedule it or scrap it.
Your inherited team will bring tremendous know-how in the business and will greatly determine your ability to get things done quickly.
In conclusion
We’re now in a world where automated AI agents can take care of much of the repetitive work that FP&A teams used to spend hours on. That shift doesn’t make your team less important. It makes the right team more essential than ever. Build around smart, creative, curious people who can adapt, solve a range of problems, and get energized by learning. The tools are here. Now it’s about working together.
