Almost every professional services firm has run an AI pilot by now, and far fewer can say what changed as a result. That gap is one of the most interesting things happening in services. It isn't really between firms that adopted AI and firms that didn't, since most have adopted AI already. It's between firms where AI produced a measurable operating change and firms where it produced a series of impressive demonstrations. The firms seeing real net human gain, meaning time and capacity actually returned to people, tended to fix where their delivery data lives first.

The broader picture looks similar. McKinsey's 2025 State of AI survey found that nearly two-thirds of organizations haven't yet begun scaling AI across the enterprise, and just 39% report any EBIT impact from it, with most of those attributing less than 5% of EBIT to AI.

The Wrong Surface

AI in services keeps getting deployed against the wrong surface: summarizing documents, drafting proposals, and tidying up notes. All of that is genuinely useful, but it sits at the edge of the business rather than in it.

The work that would actually move the numbers is different. It's knowing a project is trending toward overrun before it does, seeing a client relationship deteriorate in the communication pattern rather than in the escalation, and understanding where the resourcing constraint will form two months out. That work needs a model that can see delivery as it currently is.

In many services organizations, delivery isn't in one place. It's spread across a PSA, a CRM, a time tool, a project tool, and a set of inboxes, so the AI gets pointed at an export, and an export is a photograph of something that moves. That's a big part of why so many pilots impress and so few scale. The model was rarely the limitation, the data was.

What the Firms Seeing Gains Did First

They consolidated the delivery data model before applying intelligence to it, bringing projects, resources, time, billing, and client communication into one governed place, and only then adding the AI layer.

That order sounds obvious written down and often isn't what happens, because consolidating a data model is unglamorous and buying an AI tool isn't. But the sequence shapes the outcome. Intelligence applied to fragmented data produces confident answers about a version of the business that no longer exists.

The expectations are high. Thomson Reuters' 2025 Future of Professionals report found respondents expect AI to save professionals about five hours a week over the next year, roughly $19,000 in value per person annually, while 4 in 10 organizations are adopting AI without a strategy. That prize is realistic when the underlying data is in a state AI can use, and much smaller when it isn't.

What Changes Once the Data Is in One Place

Cloud Coach puts the intelligence along with the consolidated data rather than beside it. Its AI runs natively on Agentforce, on live records inside the firm's own Salesforce org, under the permissions the admin already maintains, so the output reflects what's actually in the system of record and can be acted on rather than double-checked.

1. Project Copilot: It answers questions about project health from live Salesforce records rather than a synced copy.

2. AI Meeting Manager: It turns a client conversation into tracked decisions and next steps on the project record, so what was agreed survives the meeting.

3. Customer Signal Intelligence: It reads the communication patterns that often come before an escalation, while there's still room to act.

None of those works well against an export, and all of them become ordinary once the data is in one place. That also changes the question worth asking any AI vendor: where does the AI actually run, and on what data? For a services business, that data includes client communications, commercial terms, and staffing decisions, so the answer is a big part of whether it's a capability or a pilot.

One Cloud Coach customer reports redeploying about 3,400 hours a year to client value, time that used to disappear into manual time entry. That's not an abstract automation claim. Those hours came back by removing a manual step from a system that already held the data.

Why Now

Plans for next year across the sector are being written on an assumption of AI leverage: headcount held flat, output up, and margin recovered. That assumption is reasonable for firms whose delivery data is consolidated, and optimistic for everyone else. The firms that get the gain are likely the ones doing the unglamorous part now, before the plan depends on it.

How We See It

Net human gain isn't a model choice, it's a data location choice, often made a year or more before the gain shows up. Consolidate first, and the intelligence becomes the easier part.

Frequently Asked Questions Related to Professional Services

• Why do AI pilots in professional services firms fail to scale?

Often because delivery data is spread across several systems, so the AI works from exports and synced copies rather than live records, and its answers can't be trusted enough to act on without checking.

• What should professional services firms do before adopting AI?

Consolidate the delivery data model first, bringing projects, resources, time, billing, and client communication into one governed system, then apply AI on top of that live data.

• Where should AI for professional services run?

Ideally inside the system of record that already holds delivery data and its permissions, so client communications, commercial terms, and staffing decisions don't need to be exported to a separate tool.

Sources Cited

McKinsey & Company, "The State of AI in 2025: Agents, Innovation, and Transformation," McKinsey Global Survey, November 2025, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

Thomson Reuters, "The AI Adoption Reality Check: Firms With AI Strategies Are Twice as Likely to See AI-Driven Revenue Growth," Future of Professionals 2025 press release, June 2025 (global survey of 2,275 professionals). https://www.thomsonreuters.com/en/press-releases/2025/june/the-ai-adoption-reality-check-firms-with-ai-strategies-are-twice-as-likely-to-see-ai-driven-revenue-growth-those-without-risk-falling-behind