System of Record vs System of Action: The AI Shift for CRMs

Running a consulting firm got more administrative in the last two years, and the numbers behind that are hard to ignore. Across firms on our platform, the average firm signed more than three times as many projects in 2025 as it did in 2023, while the average budget behind each one fell by more than half. That is more work arriving in smaller and faster pieces, and it lands as more overhead on the same operators. At the same time, a debate has taken over enterprise-software commentary. Vendors and analysts are calling it the shift from a "system of record" to a "system of action," the argument that software should stop just storing what happened and start helping you do something about it. The shift is real. Every version of the story so far, though, has been written for sales teams and their CRM. But what does this shift mean for consulting firms?

What "system of record" and "system of action" mean

What does this mean for CRMs?

For a high-velocity sales team, the pitch is simple: teach the CRM to draft the follow-up, score the lead, and nudge the rep when a deal goes quiet. A consulting firm's CRM does a different job. Most of its work comes from existing clients, so its CRM is built around proposals and live delivery capacity rather than a funnel of cold deal stages. That is why choosing a CRM for a consulting firm is its own problem, and why the sales-team version of this shift mostly solves one these firms do not have.

The bigger catch is that the CRM is only one node in the record. Resourcing, utilization, forecasting, project financials, and invoicing are what decide whether a firm makes money, and our own research shows that firms now sign three times as many projects as in 2023 at less than half the budget each. That record, far more than the CRM, is the one professional services automation software was built to run. If any record should learn to act, it is that one.

What does a system of action look like inside a consulting firm?

Picture the operational record doing the newer job. Instead of logging what your consultants did last week after the fact, it reads the same data forward. When a new project is about to land, it can look at who is free, who has the right skills, and where each person's utilization is heading, then propose a staffing shape a manager can accept or overrule. If two projects both want the same senior architect in March, it can show the clash while there is still time to solve it. And when a project's margin starts drifting from plan, it points to the reason before month-end does.

None of that removes the human; it changes what the human starts from. Today the resourcing lead walks into the Monday meeting with a spreadsheet and their own memory of who is doing what. A record that can act walks in with the answer already drafted, so the meeting is spent judging the call rather than assembling the facts.

Should the action layer sit on top of your PSA, or inside it?

There is a second version of this shift going around. The pitch is that you leave your PSA as the passive system of record and buy a separate AI orchestration layer that sits above it and does the acting. Keep the filing cabinet, and add a smart layer on top.

The problem is that this re-creates the exact tax that made the old model slow. An action layer that lives above the record has to reach down into the record for its data, through an integration, on a sync. The moment the action and the record live in two systems, you inherit the gap between them. The sync lags, fields drift, and the layer ends up acting on a copy of the truth that is a few hours or a few schema changes out of date. A confident recommendation built on stale data is worse than none, because someone will act on it.

Enterprise software already learned this the hard way. A generation of companies bolted analytics and workflow tools on top of systems like Salesforce, SAP, and Workday, then spent years and budgets keeping the copies in sync. The lesson was that an action is only as trustworthy as its distance from the record, and zero distance wins. For a consulting firm, the action that proposes a staffing move or points to a margin risk has to be built on the same governed data it comes from, not a mirror of it sitting one integration away. Record and action belong in one place, on one source of truth.

Why the operational record still has to be right first

All of this comes with a condition that is easy to skip past, so here it is plainly. A record that acts is only as good as the data underneath it. Point an advisory layer at incomplete timesheets, half-updated allocations, and forecasts nobody has touched in a month, and it will hand you wrong answers with total confidence. Run it on bad operational data and it amplifies the mistake faster than any passive filing cabinet ever could, because now the mistake ships as a recommendation someone trusts.

That is not an argument for waiting. The move is to get the record right first, and firms already know it. When we asked professional services firms what they most wanted to improve in how they win work, data integration and accuracy came second, named by 41.2% of them.

How to evaluate this shift for your firm

So how should a firm leader judge any of this when a vendor pitches it? A few questions cut through the noise. Start by asking what the AI is acting on: is it reading your real operational data where it lives, or a synced copy that is already drifting? The closer to the source, the more you can trust what it tells you. Then ask what it actually does. Does it propose a specific action you can take, or just summarize what already happened in prettier words? A summary is a report with better manners, while an action changes what you do next. Finally, ask who approves. The honest answer from anyone credible today is that a human does, every time. The best of this technology right now is advisory, which means it recommends and the operator decides. Anyone selling autonomous decisions for something as consequential as staffing or margin is selling ahead of what the technology has earned.

Advisory is not a weakness to apologize for. It is exactly what a firm should want while the record is still earning that trust.