Industry
Professional Services & Consulting: what to automate, what to keep human.
AI in professional services should give principals and PMs their billable hours back. The non-billable layer, scheduling, document wrangling, proposal boilerplate, status reporting, quietly eats partner capacity no one bills for, and taking even half of it back is real revenue per head, per month. Most vendors sell the opposite project, AI that drafts the deliverable, which mostly hands a principal a confident document they still have to re-derive to put their name on. Aim AI at the admin and the billable judgment hours come back. Those are two different projects, and most firms keep buying the deliverable-drafter while needing the hour-reclaimer.
The argument
AI in professional services should give principals their billable hours back. Vendors sell deliverable-drafters that mostly hand a partner a document they must re-derive to trust. AI pointed at the non-billable admin, scheduling, document wrangling, proposal boilerplate, status reporting, is what actually unlocks billable judgment hours, and that math compounds per partner, per month.
Where the billable hours are actually going
The early money in services AI is in the non-billable layer, and the data is on it. In Bluebeam's 'Building the Future: AEC Technology Outlook 2026' (1,000+ technology decision-makers, October 2025), early AI adopters are showing hard ROI: 68% saved at least $50,000 and 46% reclaimed 500 to 1,000 hours, and Bluebeam's own report says those reclaimed hours came from 'scheduling, planning and document analysis', not from drafting the deliverable. That is exactly the surround a partner-led firm pays its highest-billing people to do for free.
The leak in a professional-services P&L is rarely 'we draft reports too slowly.' It is that principals and PMs spend a large share of their week on work no client is billed for: scheduling, proposal boilerplate, document chasing, status updates, internal reporting. Take half of that back from named senior people and it converts straight into capacity at the principal rate. That is real revenue per partner per month, and it is the project most firms never separate out from the louder one being sold.
So a serious firm's question is not 'can AI write the report.' It is 'where are my principals' hours leaking, and is that where this tool is actually pointed.'
Where to deploy AI first in a services firm
Move first on the non-billable surround, where being wrong is cheap and a principal is not putting their name on the output. Proposal boilerplate and past-project assembly a partner edits. Scheduling and document chasing across a project. Meeting notes turned into a status update or an internal SOP. First-draft client reporting from data the firm already trusts. None of these is the deliverable, and each one hands time back to someone billing at the top rate.
These earn the harder project, AI assisting the deliverable itself, once a review discipline exists: a named principal who owns the reasoning, not just proofreads the prose. Reclaim the non-billable hours first, then earn the billable one.
Why drafting the deliverable is the wrong project to buy first
There are two completely different AI projects inside a professional-services firm, and treating them as one is the mistake. Project one: AI drafts the billable work product. Project two: AI removes the non-billable work that surrounds it. They look adjacent. They are not the same purchase, and only one of them reliably converts to margin without new review burden.
Adoption is still low: Monograph reports only 6% of architects regularly use AI in their work, and Bluebeam found only 27% across AEC use it at all, so most firms are early enough to choose which project they run first. AI that drafts the deliverable attacks billable time, where a human has to re-own the output anyway, because the deliverable carries the firm's name, its professional liability, and the judgment the client is paying a principal for. An AI draft is not a finished deliverable, it is a confident document a principal still has to reconstruct the reasoning of to defend it. Fast to draft, not fast to stand behind.
Practically: separate the two projects on paper before buying. Name the non-billable work AI removes from named senior people, with the hours it is eating. Keep the deliverable human-owned: AI may assemble inputs and prepare backup, but the reasoning, the recommendation, and the signature stay with the principal whose name is on it. Buy the hour-reclaimer first.
The hour-reclaim plan, and the project not to buy first (use this one)
This is the artifact. Not 'adopt AI thoughtfully': the one page a firm's partners fill in before buying anything, so the hours coming back are real billable capacity and the deliverable-drafter does not get bought before there is a review discipline that can catch a confident wrong draft. Nothing is purchased until each line is true:
- The reasoning and recommendation in the deliverable. AI may assemble inputs and backup; the judgment and the signature stay with the principal who carries the liability.
- Anything that goes to a client under the firm's name without a named principal having re-owned the logic, not just read the prose.
- Scoping and pricing an engagement. That's a risk judgment, not a drafting task.
- The client conversation where the recommendation is delivered and defended. That is the service.
- Which of the two projects goes first. Buying the deliverable-drafter before the hour-reclaimer is the default mistake; it's a partner decision, made deliberately.
How you'll know it actually worked
Measure services AI by hours, both kinds. Non-billable admin on named senior people: did it drop, and did the reclaimed time show up as billable utilization at the principal rate. Billable: did the deliverable get faster without principal review and rework hours rising to match. If non-billable goes down and billable utilization goes up, the hour-reclaimer worked. If drafts get faster but principal review hours climb because every confident draft has to be re-derived to be signed, you moved time from drafting to defending and called it productivity. That shows up in utilization, not in the demo, about a quarter after rollout.
How ADA helps professional services & consulting
Service paths
AI Implementation for Professional Services & Consulting
Partner- and principal-led professional-services firms (consulting, advisory, small/mid AEC, 10–200 people) where the highest-billing people lose a large share of the week to non-billable admin and the board wants an AI answer.
View service ➞Workflow Automation for Professional Services & Consulting
Firms whose principals and PMs run on remembered admin: proposal chasing, status updates, document handoffs, internal reporting, and who want the repeatable surround to run without anyone automating the judgment.
View service ➞Related workflows
Frequently asked
Should a professional-services firm use AI to draft deliverables?
Not first, and not without a review discipline. The deliverable carries the firm's name and liability, so an AI draft still has to be re-derived by a principal to be signed: fast to produce, not fast to stand behind. Move the non-billable admin work off senior people first; bring AI into the deliverable only once a named principal owns the reasoning and the AI never supplies the judgment.
What should consulting or A/E AI never do on its own?
Supply the reasoning, the recommendation, or the signature on anything that leaves under the firm's name; scope or price an engagement; or run the client conversation where the recommendation is defended. It assembles inputs and clears admin; the judgment the client pays for stays human.
What's the first AI workflow a services firm should actually build?
One in the non-billable surround: proposal boilerplate assembly, scheduling and document chasing, meeting-notes-to-status, or first-draft internal reporting from trusted data. Being wrong is cheap there, no principal signs it, and every hour reclaimed converts to capacity at the top billing rate.
How do we know if AI actually improved the firm's economics?
Non-billable hours on named principals and PMs drop and billable utilization rises, not deliverable draft speed. If drafts speed up but principal review hours rise to match, the time moved from drafting to defending and no capacity came back.
Sources
- Bluebeam: 'Building the Future: AEC Technology Outlook 2026' (1,000+ technology decision-makers, Oct 28, 2025): only 27% of AEC firms use AI; among adopters 68% saved $50,000+, 46% reclaimed 500–1,000 hours, from scheduling, planning and document analysis; 94% of adopters will increase usage in 2026
- ASCE: 'Architecture, engineering, construction sector slow to adopt AI, survey shows' (Dec 18, 2025), reporting the Bluebeam survey: 27% currently use AI, 94% of those will increase usage in 2026
- Monograph: 'AI for A/E firms': only 6% of architects regularly use AI for their work; utilization benchmarks drawn from 527 A&E firms
Next step