55% of Employers Who Cut Staff for AI Regret It — What That Means for Your Firm
Published January 6, 2026 · By The Crossing Report
55% of Employers Who Cut Staff for AI Regret It — What That Means for Your Firm
The story you hear most often in 2026 is the one about AI eliminating roles. The one you hear less often is what happens after.
Forrester's 2026 Future of Work report tracked companies that made AI-driven workforce cuts and then checked back in. The finding: 55% regret it. Among those, 35.6% rehired more than half of the workers they let go. One in three spent more on restaffing than they originally saved.
This isn't an argument against AI. It's a data point about sequencing — and it matters especially for professional services firm owners who are weighing whether to let someone go because an AI tool can theoretically do the job.
What Actually Happened
The companies in the Forrester data didn't set out to make a mistake. They saw compelling vendor demonstrations. They ran pilots. They read the same research projections about AI's efficiency gains. Then they cut.
The problem, which Forrester's analysts identified across dozens of case studies, is that most of these firms cut based on AI capabilities that weren't yet production-ready — or that required significantly more human oversight than the demo suggested. The gap between "AI can do this in ideal conditions" and "AI can do this reliably on your actual client work with your actual data" is still wide in 2026 for most professional judgment tasks.
One in three firms ended up spending more on restaffing — finding, hiring, and onboarding replacement workers — than they'd saved by cutting. That doesn't account for the revenue disruption during the gap, the institutional knowledge that walked out the door, or the morale impact on the team members who remained.
Why This Pattern Is Especially Relevant for Professional Services
If you own an accounting firm, law firm, consulting practice, staffing agency, or marketing agency, the Forrester data applies to your situation differently than it applies to a retail chain or a call center.
In professional services, the work that makes you money is the judgment work — the advice, the analysis, the client relationship, the professional sign-off. AI is currently most reliable at automating the support tasks surrounding that judgment: transcribing meetings, drafting first-pass documents, coding transactions, flagging contract deviations, scheduling, generating reports.
The firms that are regret-free in the Forrester data are predominantly the ones that used AI to expand capacity without cutting headcount — meaning they let AI handle high-volume, low-judgment tasks so their existing people could do more of the judgment work. Revenue went up. Costs didn't significantly change. Margins improved.
The firms with regret cut people first and then discovered that AI's handling of client-facing or judgment-intensive work was unreliable enough to require the kind of human oversight that either negated the savings or produced errors they had to fix.
Three Workforce Scenarios and How to Think About Them
Scenario 1: You're thinking about not replacing someone who left.
If a support-level employee leaves — admin, data entry, document processing — and you're wondering whether AI can cover that function before you hire, this is the lowest-risk version of the question. AI handles transcription, scheduling, first-draft documentation, and transaction coding at a level of reliability that makes a deliberate trial reasonable. Run the pilot before posting the role. Give it 60-90 days. If AI covers 80%+ of the function reliably, you have real data for the headcount decision.
Scenario 2: You're thinking about cutting someone in a role that involves client contact.
This is the higher-risk version. Client-facing roles — account management, advisory, intake, relationship management — are where the Forrester regret cluster is highest. AI can draft client communications, but the relationship layer requires a human. The calculation for cutting here requires answering: what percentage of this role is genuinely automatable vs. judgment-based? If you can't quantify that clearly, the Forrester data suggests waiting until you can.
Scenario 3: You're thinking about restructuring a team around AI leverage.
This is the version the data supports. Instead of cutting headcount, you add AI to each person's workflow — automating the tasks that consume the most hours but require the least judgment — and track what each person can now do with the recovered time. A 5-person team doing the work of 7 is a very different proposition than a 5-person team with 2 empty chairs and AI coverage that isn't quite there yet.
The Vendor Promise Gap
One reason the Forrester regret rate is as high as 55% is the vendor promise gap — the distance between what AI tools are demonstrated to do and what they reliably do in production.
Tool vendors, understandably, show their best cases. The demo environment has clean, structured data, cooperative documents, and controlled inputs. Your practice has legacy file formats, non-standard client data, edge cases, and exception workflows that the demo never touched.
This doesn't mean the tools don't work. It means the pilot environment is where you discover the edges of the tool's reliability — and that discovery needs to happen before the headcount decision, not after.
For professional services firms: run any AI tool on real work for 90 days before letting it drive a staffing decision. Track the error rate on actual client documents, not the accuracy metric from the vendor's benchmark dataset.
The Right Question to Ask Right Now
The question isn't "can AI replace this person?" It's "which tasks in this person's workflow can AI handle reliably enough that I don't have to assign those tasks to a human anymore?"
The answer to that question is almost never 100%. It's more often 40-60% for an administrative or support role, and 15-30% for a client-facing or advisory role. That leaves a gap — and that gap is where the judgment work lives, which is usually the highest-value work the person is doing anyway.
Start there. Automate the tasks. Track what the person can now do with recovered hours. That's the version of the AI workforce question that produces results without a regret cycle.
This week, by firm type:
- Law firm: Identify the three most time-consuming administrative tasks any attorney or paralegal does that don't require professional judgment (scheduling, document formatting, transcription). Assign one to an AI tool this week. Track the time recovered.
- Accounting firm: Run your meeting notes through Fathom for the next two weeks before deciding whether to reduce admin support. The actual time savings data you collect is worth more than any projection.
- Consulting firm: Map which deliverable prep tasks (data organization, literature synthesis, slide formatting) AI can draft reliably. Give your team two weeks of AI-assisted prep before any headcount conversation.
- Staffing agency: Track how many hours per week your recruiters spend on tasks that AI tools (Loxo, Bullhorn AI scoring) can now handle. Build that baseline before making any staffing decisions about your own firm.
- Marketing agency: Identify the recurring deliverable that consumes the most team time. Run it through AI for 30 days. The error rate you observe is the input to any workforce decision — not the vendor data sheet.
The Forrester data says the regret comes from moving faster than the reliability data supports. Slow down one month and get the real numbers.
Related reading: The AI Adoption Gap in Professional Services: Where Firms Are Actually Stuck | Your Firm's AI Governance Gap Is a Liability, Not a Best-Practice Gap
Frequently Asked Questions
What does the Forrester 2026 Future of Work report say about AI-driven layoffs?
Forrester's 2026 Future of Work report found that 55% of employers regret AI-driven layoffs. Among those, 35.6% rehired more than half of the workers they let go — and one in three spent more on restaffing than they originally saved by cutting. The underlying problem: most of these firms cut based on AI capabilities that weren't yet production-ready or that required more human oversight than expected.
Should professional services firms cut staff to replace them with AI?
The Forrester data argues for caution. In professional services — law, accounting, consulting, staffing, marketing agencies — the work that AI handles reliably (document drafting, data entry, transcription, scheduling) is typically support-level work, not the client-facing judgment that drives revenue. Cutting client-facing staff to 'save on salaries' while AI handles client work is a different and riskier calculation than automating administrative tasks while keeping your delivery team intact.
What's the difference between automating tasks and replacing people with AI?
Automating a task (having AI draft a first-pass contract review, transcribe a meeting, or code transactions) adds capacity without changing headcount. Replacing a person with AI means eliminating a role entirely and trusting AI to handle everything that person did — including the judgment calls, client relationships, and exception handling. The Forrester regret data suggests firms that moved fast on the second approach found AI couldn't yet reliably handle the full scope of the role.
What should professional services firm owners do instead of cutting staff for AI?
Start with task automation, not headcount cuts. Identify the highest-volume, lowest-judgment tasks in your workflow — meeting transcription, first-draft client communications, transaction coding, contract flagging — and automate those first. Let AI free up your existing team's capacity before making any workforce decision. If the math still points toward headcount reduction after 6-12 months of automation, you'll have real productivity data to make that decision on — not vendor promises.