Hackett Group ran its Q2 call after the close on August 4. The company sells AI transformation consulting for a living. For the past year it has been running that transformation on itself, and this is the first quarter with a clean year-over-year comparison on both sides of the ledger.
Start with the line the release leads on. GAAP diluted EPS of $0.18, against $0.06 in Q2 2025. Tripled.
Now the line that did not make the headline. Adjusted diluted EPS of $0.34, against $0.38 a year ago. Down 11%.
The tripling is arithmetic against a depressed 2025 base. The number that describes the operating business is the second one, and it is moving down.
The unit math
Revenue before reimbursements came in at $68.3M, against $77.6M in Q2 2025. Down 12%.
Over the same stretch, consultant headcount went from 1,382 to 1,211, a loss of 171. Total headcount went from 1,685 to 1,491, a loss of 194, or 11.5%.
Divide one by the other.
Q2 2025: $77.6M across 1,382 consultants is $56,154 per consultant for the quarter. Q2 2026: $68.3M across 1,211 consultants is $56,400. A gain of 0.4%.
Run it against total headcount and it goes the other way, from $46,053 to $45,808. Down 0.5%.
A firm that sells AI productivity as a product spent a year applying it in-house and produced no measurable change in output per person. Revenue fell 12%, headcount fell 12%, and the two lines walked down together. What Hackett banked was payroll, not throughput.
Severance with a strategy label
The company is unusually plain about the mechanism. It booked roughly $500K in “AI transition charges” in Q2, and guided to about $1M more in Q3, consisting mainly of severance tied to aligning headcount with AI-driven productivity.
Severance is severance. Filing it under a line item called AI transition changes what an investor sees: not a contraction, an investment with a direction. The whole sector learned this move this year. In Challenger’s July report, AI led the stated reasons for layoffs for a fifth straight month, and Andy Challenger’s read was blunt: naming AI in a layoff announcement wins over investors while pushing current and prospective employees away.
Set the quarter’s cash flows side by side. Hackett repurchased $4.0M of its own stock in Q2, 372,000 shares at an average of $10.56, and paid out $3.0M in dividends. That is $7.0M returned to shareholders against roughly $500K in severance-driven AI transition charges. Shareholders collected 14 times what the exits cost.
What makes Hackett worth the attention is not that it did this. It is that the firm is small enough and discloses finely enough to break out consultant headcount as a separate figure. Most of its peers do not publish that number, which is precisely why nobody can run this division on them.
The order is backwards
The most important sentence on the call came from CEO Ted Fernandez, who said the firm is currently capturing about 20% of the available productivity gains and is targeting 50% by year end.
Read that against the headcount. The 194 departures already happened. The 50% is guidance.
If the productivity lands, this was harvesting early. If it does not, this was an ordinary cost cut wearing a better coat. The test date is specific: the Q4 report in February 2027, and the question is whether revenue per consultant has left the $56K mark it has now held for four quarters.
The segment detail says where the money actually came from. Global S&BT fell from $43.6M to $35.6M, with contribution down from $13.0M to $9.1M. Oracle Solutions fell from $20.5M to $15.3M. Only SAP Solutions grew, from $13.5M to $17.4M, with contribution up from $3.9M to $5.6M. What held the quarter together was SAP implementation work, not the AI platform line.
The platform line is not empty. Hackett closed more than $30M in platform-led wins late in the quarter, ramping through Q3, some of it running to the end of 2027. Those are real contracts. They are also in the future. The 194 people are in the past.
The Q3 guide deserves its own look. Revenue before reimbursements of $68.0M to $70.0M, adjusted EPS of $0.37 to $0.39. Fernandez calls the quarter an inflection point. The top of that range, $70.0M, still sits below the $77.6M the firm posted a year ago. The inflection is sequential. It is not a return to the old line.
Consulting is the leading indicator for white collar
Consultancies run the experiment before their clients do. Their cost base is almost entirely people, with no plant and no inventory to absorb a decision, so any judgment about AI efficiency shows up in headcount within two or three quarters rather than two or three years.
The curve has passed several markers already. Accenture booked $865M on a reskill-or-exit program and reached 85,000 AI and data specialists by March. PwC cut about 600 executive assistants in February and McKinsey about 200 late last year, taking out exactly the layer their internal assistants handle best. Hackett, at 1,491 people, turns faster than any of them, and it has now traced the full shape of the curve: cut proportionally, hold output per head flat, book the productivity as next quarter’s guidance.
If you work in consulting or professional services, there is one usable test in here. Watch whether your firm reports total revenue or revenue per head. Per-head is the only figure that can falsify the AI productivity story, and it is the one most firms decline to break out. Hackett broke it out. It came back flat.
