AI leadership in professional services: Summary & key takeaways
The data: 88.9% of services firms rate their leadership as supportive of AI, up from 75.0% in 2025 (SPI Research, 2026 Impact of AI on Professional Services).
The catch: That support lines up with better delivery, but not with profit.
The answer: Build what SPI says turns support into results: a scorecard with an owner, ready data and trained people.
The payoff: With those in place, you can price the outcome, not the hours.
Agreeing with an idea is easy. Doing it is the hard part. I recently attended a workshop on outcome-based pricing at TSIA World in Las Vegas, where everyone agreed with the logic. Yet outcome-based pricing is just 3% of services revenue on average, according to TSIA, and the median is 0%.
AI leadership has the same gap. In SPI Research's 2026 Impact of AI on Professional Services report, which we sponsored, 88.9% of 245 firms rate their leadership as supportive or very supportive of AI. Yet that support doesn't reliably show up in profit.
So what does leading AI involve? SPI found that firms that turn support into results pair it with three things: a scorecard, ready data and trained people. This article covers how to build each, and how they let you price the outcome.
Support lines up with delivery, not profit
SPI asked firms to rate how strongly their leadership backs adopting AI, from very unsupportive to very supportive. Here's how each group performs.
)
In SPI's data, firms with very supportive leadership deliver 83.0% of projects on time, against 73.0% where leadership is neutral or less supportive. Their revenue also grows faster, at 7.9% against 1.3%, though SPI hasn't confirmed either gap as reliable once firm size and type are accounted for. Profit is a different story: EBITDA is 17.3% for very supportive firms and 17.9% for neutral or less supportive ones.
That's why SPI calls support necessary but not enough. Here's what to build alongside it, starting with the piece a leader can set up this week.
1. Give AI a scorecard and an owner
SPI's advice to leaders is specific: give AI a named owner, a budget and a scorecard, and review it as often as you review utilization and margin. Every initiative should earn its place against four goals: growth, your people, the client experience and profit. If you can't name the number it should move, hold the budget.
At TSIA World, we shared five numbers to track per engagement. They are what work agents did, what each agent cost and who owns it, human review time, rework rate and margin per outcome.
The owner needs a view of the whole engagement, because that's where margin goes missing. Sales, delivery and finance each hit their own goals, while a higher-cost stand-in, one more client change or a two-week onboarding slip takes the margin. Most firms only see the damage at invoicing. For why saved hours don't turn into margin on their own, see AI shrinks the work, not your overhead.
We're building Teamwork.com, the agentic PSA, to give owners that view. Resourcing and financials are captured as delivery happens, so you can track project profitability live.
2. Start where your data is ready
A scorecard is only as good as the data behind it. SPI ranks AI use cases by value and data readiness. The safest starting points are drafting deliverables, proposals and RFP responses, contract review and meeting summaries. Planning and pricing analysis can wait until your PSA, CRM and financial data can be trusted.
I apply the same rule to my own team. I vibe code side tools and dashboards for marketing: I describe what I want in plain English, and AI builds it in hours. When one breaks, it can sit for a week or two, which is fine for a side tool. It isn't fine for the systems you run the business on.
A simple test: if a tool broke for two weeks, would a client, an invoice or your numbers be affected? If so, it needs a dedicated team behind it.
3. Train people on the outcome, not the tool
Data is half the foundation. The other half is people who know what the tools are for, and SPI's advice is to make AI proficiency universal, starting with training.
When I was at LinkedIn, ads had a big early churn problem. Advertisers expected Google or Meta results, and agencies reported likes and clicks, so LinkedIn looked poor even though its long-term ROI proved higher for B2B. What helped was LinkedIn Labs, an online course that led with how to get your outcomes, then how to use the platform.
AI training works the same way. Start with the outcome people care about, such as a faster first draft, then show them the tool that gets them there.
The payoff: price the outcome
With those three in place, you can do what most firms still can't: price the outcome. TSIA sets out three tests first: what the work costs to deliver, whether you and the client agree when it's done, and what the client spends on it today. Your scorecard and data answer the first. The other two are a conversation with your client.
Start that conversation on value, not hours. Clients can judge AI-enabled work by the time it took, the way advertisers judged LinkedIn by clicks. I hear it in customer and prospect conversations: many now expect lower rates. But if work that cost $100 now takes $10 of billable time, the client still gets $100 of value.
So ask what the outcome is worth to them. Then agree the real outcome ("more leads" and "more pipeline" are very different asks), how it's measured and how you'll report on it.
We buy this way ourselves. Our event agencies get a set amount to bring a set number of attendees who fit our ideal customer profile, not a flat fee per event. We're happy to pay a little more for better quality.
The evidence on these conversations is still thin, so treat this as a starting point. For the rates SPI found, see AI-first firms charge more, not less.
Where to start this quarter
Everyone at that workshop agreed. Agreement was never the hard part. This quarter, name the owner and pick one safe use case. Then ask these questions about your three largest clients:
Who owns margin on this engagement, from quote to invoice?
What outcome are we paid for, and how is it measured?
Which systems hold this client's numbers, and is there a team behind them?
Wherever you can't answer, start there.
)
)
)
)
)
)
)
)
)
)