PS AI maturity model: five stages from spreadsheets to agent-operated

What is the PS AI maturity model?

The PS AI maturity model describes five stages of how a professional services firm runs its operations: manual, connected platform, AI-assisted, agent-driven and intelligence-driven. Each stage builds on the data of the one before. It is a diagnostic: it shows where your operations actually are, what to build next and which AI investments pay off at that stage.

Stage 1: manual operations

Operations run on spreadsheets, email and tools that don't share data. Staffing happens in a spreadsheet, margins are calculated at month end, timesheets are chased on Fridays and approvals travel by email. It works while the delivery lead knows everyone personally, and stops working as the firm grows.

The stage is about infrastructure, not size: fast-growing firms and firms built through mergers are often here too. When a key operations person leaves, their knowledge leaves with them.

What to build next: one system that connects pipeline, staffing, time, project financials and invoicing. The goal is data, not intelligence.

Stage 2: connected platform

A PSA connects the operational lifecycle: pipeline, staffing, time, project financials and invoicing run through one system, and dashboards show what happened. The data is connected but passive. Every staffing decision, margin review and approval still needs a person to open the tool, read the data, decide and act.

The platform is the system of record; the operations team is still the operator. Choosing and rolling out a PSA gets a firm here, not further. If you are choosing one, our guide to the best PSA software groups 12 tools by firm type and how far their AI goes.

What to build next: live visibility instead of monthly reports, alerts when thresholds are crossed, and the data quality later stages need: complete time entries, accurate allocations and consistent rates.

Stage 3: AI-assisted operations

AI tools sit on top of the connected platform. Assistants answer data questions, copilots draft proposals and reports, and suggestions fill timesheets and staffing shortlists. People work faster, but they still make every decision, so operations headcount still grows with the firm. A firm can call itself AI-powered at this stage without changing how it operates.

The test: does the AI do the work, or help a person do it faster? At this stage, it helps.

What to build next: pick one operational job, staffing is usually the one with the most impact, and let an agent do it within rules you set, rather than only recommending.

Stage 4: agent-driven operations

Agents do core operational work within rules people set: proposing teams for open roles, keeping skills and project data current, pre-filling time cards and flagging projects that are losing margin. People review, approve and handle exceptions, and the operations team owns the rules: what agents may do alone and what needs approval.

What changes at this stage:

• Staffing proposals arrive ready to review instead of being built from scratch.
• Margins are watched continuously, not at month end.
• Timesheets start pre-filled, so they are complete more often.
• Operations work no longer grows in step with headcount.
• Every agent action creates data that the next stage builds on.

This is the stage an agentic PSA is built for. For what agents take over inside the firm and how to start, see AI agents in professional services.

What to build next: a habit of recording outcomes, such as margin, delivery quality and client satisfaction per engagement, so that decisions can learn from results.

Stage 5: intelligence-driven operations

Decisions draw on what worked before: which team compositions delivered the best margins on which project types, which early signals predicted a project going off track, which human-to-agent ratio suited which phase. Agents use those patterns when they propose teams, prices or scope changes.

This stage needs years of clean operational and outcome data, so it starts with your own history, and no firm can skip the stages that produce that data.

What to build next: keep the loop running. Each engagement's outcome data improves the next proposal.

How do you assess your firm's AI maturity?

Answer the questions below for your operations as they run today, not for the tools you own. Your stage is the highest one where you can answer yes to every question, and the stages above it show what to build next. Be strict: owning a tool doesn't count as using it.

Stage 2: connected platform

• Do pipeline, staffing, time, project financials and invoicing live in one system?
• Can you see availability and project margins without building a spreadsheet?

Stage 3: AI-assisted

• Can people ask questions about utilization, availability or margins and get answers in seconds?
• Do consultants start from suggested time entries instead of empty timesheets?

Stage 4: agent-driven

• Does at least one operational job, such as staffing proposals or data upkeep, run through agents, with people approving?
• Are the rules written down: which agent actions need approval and who can change that?
• Can you see what agents changed, and undo it?

Stage 5: intelligence-driven

• Do you record outcomes per engagement, such as margin, quality and client satisfaction?
• Do staffing and pricing proposals draw on those outcomes?

If you can't answer yes to the stage 2 questions, you are at stage 1, whatever AI tools you have bought.

How should you use the maturity model?

Use it to choose the next step, not the final one. Build the stage directly above where you are, judge platforms by whether they carry you to the stage you are building toward, and track one transition metric per stage, so progress shows in the numbers rather than in tool purchases.

• Build the next stage, not two ahead. A stage 1 firm needs a connected platform, not agents. A stage 2 firm needs visibility and clean data, not benchmarks. Skipping stages leaves agents working on data nobody trusts.
• Judge platforms by your target stage. A platform built for reporting carries a firm to stage 2. Ask whether its agents do operational work today and whether you can see and approve what they do.
• Track one metric per transition: stage 1 to 2, the time from a deal closing to its first time entry; 2 to 3, the time spent looking up data by hand; 3 to 4, the share of staffing proposals made by agents; 4 to 5, the share of decisions that draw on outcome data.

Agileday is an AI-native professional services automation (PSA) platform. Its AI agents do the manual work across operations, such as building optimal teams, keeping data up to date and analyzing deals, while people make the decisions. To place your firm on the model with someone who runs these operations every day, talk to an expert.