The way consulting firms staff projects has not fundamentally changed since the 1990s. One human, one role, one rate card. The tools got better – spreadsheets became PSA platforms, email became Slack – but the operating model stayed the same: estimate the hours, assign the people, bill the time.
That model is starting to crack, and for the first time, Salesforce customers have a credible way to change it. Agentforce makes it possible to put an AI agent on a project the same way you put a person on a project.
This article walks through what that looks like in practice: where the pressure is coming from, what an agent actually does on an engagement, why governance is the part most people skip, and how project economics shift when you get the human-to-agent mix right.
The Numbers Say the Old Model Is Cracking
The 2025 SPI Professional Services Maturity Benchmark surveyed 403 firms. EBITDA margins sit at 9.8%, the lowest in five years. Billable utilization has dropped to 68.9%, below the 75% most firms treat as healthy. On-time delivery is at 73.4%. Firms are asking people to do more, and the numbers say it is not working.
The problem is not effort, it is the shape of the work. Every engagement carries a layer of structured, repeatable work – scoping, status reports, timesheets, meeting recaps, milestone nudges – that eats consultant hours without needing consultant judgment. That work has to get done. It does not have to get done by the person you are billing out at $200 an hour for strategic thinking.
What “AI Agent as Project Resource” Actually Means in Salesforce
Picture a Monday morning standup. The project manager opens the resource plan and scrolls down the team list. Two senior consultants, a solution architect, a junior analyst, and a row labeled “Scoping Agent.” That row has a task assigned, a deadline, and a deliverable already sitting in the approval queue, drafted and waiting for review. The rhythm shifts from collecting status to making decisions.
That is what an agent on a project looks like: not a chatbot or a sidebar copilot, but a resource on the plan, built in Agent Builder, running on your Salesforce org, with assigned tasks, expected outputs, and the same visibility in reports and dashboards as the humans next to it.
The Governance Gap: Who Approves What the AI Produces?
Most AI conversations focus on generation. The harder question is governance: who reviews the output, who approves it, and what the audit trail looks like when it reaches the client. This is where most AI pilots quietly die, and where the Salesforce platform gives you something the rest of the market does not.
Every piece of agent output should pass through a structured approval step. A human reads it, edits if needed, and explicitly approves before anything moves to the client or a downstream system. The agent does not decide what ships. The human does.
Running this on Salesforce is the part that matters. Your approval flows, permissions, audit trail, and data model are all on the same platform your delivery team already uses. You are not shipping data to an outside LLM and hoping for the best, you are using the same trust and sharing model you apply to every other object in your org. Firms that skip this layer will learn why it exists the first time an unreviewed output lands in a client’s inbox.
How This Reshapes Project Economics
When agents handle the structured execution layer, consultants focus on what humans are actually good at: judgment, relationships, complex problem-solving, and the thinking that earns premium rates.
This does not mean smaller teams, it means differently shaped teams. A project that once needed three consultants at 40 hours each might now run with two consultants plus three agents handling the operational throughput. Total delivery capacity goes up, margins improve, and humans stop burning out on status reports.
Forecasting changes with it. Instead of asking “how many people next quarter?”, the question becomes “how many people plus how many agents?”. Real margin expansion will belong to firms that stop treating AI as a personal productivity tool and start treating it as a line on the resource plan.
Where Klient PSA Fits
Klient PSA runs 100% natively on Salesforce. Every project, resource, timesheet, and approval already lives on the platform.
We ship a catalog of eight specialized Agentforce agents, each built for one job in professional services delivery: scoping, project status, timesheets, meeting prep, onboarding, case resolution, knowledge capture, and development. Every output runs through a structured approval step before it reaches a client.
We are demonstrating the full hybrid project delivery model live at the Agentforce World Tour in New York on April 29 and in Toronto on May 7.
