Task speed is not firm value
AI tools saved tax practitioners an average of a third of their time on tax preparation and raised throughput by about 50% in the 2025–26 tax season, according to data Current shared with Nitin Nohria and Hélène Hannecart for their Harvard Business Review article of November 11, 2026. Their point is that this kind of speed does not automatically become an equal gain for the firm.
HBR describes Current as a platform that provides technology and shared services to partner accounting firms. The figure is a third of the time spent on tax preparation, not a third of anyone's working week.
Current has also published its own numbers. In its June 2, 2026 announcement, Current said its Tax AI processed 7,000 tax returns that season and cut time spent "by about a third," with participating firms seeing an average of 31 percent savings in tax preparation time. Those are Current's figures about its own product, not HBR findings.
One disclosure matters here. Thrive Capital announced in January 2022 that Nohria, a former dean of Harvard Business School, had joined it as partner and executive chairman. Current names Thrive Holdings as an owner-operator, and Forbes reported on June 2, 2026 that Current is owned by Thrive Holdings, a spinoff of Thrive Capital. Read the efficiency figures with that link in mind. The article's central warning is a different matter, and it is the part most useful to an owner.
HBR's summary states it plainly: "AI can dramatically accelerate individual tasks without delivering equivalent gains for the organization, because delegating work creates costs beyond execution." The useful question is not how many hours a tool saved, but what happened to them.
The bottleneck moves to review
When preparation gets faster, the constraint moves to review. That is what Nohria and Hannecart found across four accounting firms. According to the article's abstract, faster tax preparation "can shift bottlenecks to review."
Every return a tool helps prepare still needs a qualified person to check it before it reaches a client. If preparation output rises and review hours stay the same, work queues in front of the reviewers.
This is not unique to tax work. Many owners already treat AI output as something to check. Our roundup, What AI Tools Are Small and Mid-Sized Businesses Actually Using?, notes that 67% of AI-using owners in a Network Solutions survey check AI's work rather than trusting it outright. That checking takes time, and it often lands on the people senior enough to sign off.
Remote teams can miss this shift. In an office, a partner can see the stack of files waiting for sign-off. On a distributed team, the backlog sits in a shared queue or someone's inbox, and it may go unnoticed until a deadline slips. The habit we recommend in Version Control for AI Agents, a human review for each agent task, carries over to any AI-assisted work: name who reviews, and make the queue visible to the people who manage it.
Incentives and the next reviewers
Faster work can also weaken incentives and the training of the people who will review the work later. HBR's abstract says the evidence from the four firms shows how faster tax preparation can "undermine employee incentives, and disrupt the development of future professionals."
Leitgrad's November 12, 2026 summary of the HBR article spells out a mechanism the authors describe. In Leitgrad's account, the routine work that junior professionals used to learn on falls away, and when it is not clear who the gained time belongs to, incentives in the team suffer. That detail is Leitgrad's reading; the HBR Store summary names the incentive and development risks without spelling out that mechanism.
Both risks apply well beyond accounting. On incentives, if people are measured by hours or volume, a tool that shortens a task changes what a good day looks like. If nobody says what the saved hours are for, each person answers alone.
On development, routine preparation is how future reviewers learn what a wrong number looks like. If juniors stop doing that work and nothing replaces it, the business shrinks the pool of people who can handle the review that has just become the bottleneck. On a remote team, juniors already have fewer chances to learn by watching colleagues. We look at that problem in a coming article, "Your team stopped asking each other questions."
Assign the freed hours
Freed hours go wherever someone decides they go, so the owner or partner should decide, and write it down. The firms in the HBR study already used the time in different ways. HBR reports that the tools freed practitioners to spend more time with clients, and that one firm used the additional capacity to accept and complete approximately $300,000 of additional tax work late in the season.
Both are reasonable uses. The point is that they are choices. HBR's abstract says leaders must "redesign workflows and roles, and decide how to use the capacity AI creates." In a remote business, if leadership does not make that decision, it gets made person by person, out of sight.
A written plan for freed hours can fit on one page. For each team or role using AI, it should answer four questions:
- Where do the hours go? Name the priority: client service, quality, or growth. If it is a mix, say which comes first when there is not enough time for all three.
- Who reviews, and with how much time? Review capacity has to grow with output. That can mean more reviewer hours, a second reviewer, or a deliberate cap on throughput until review catches up.
- How do juniors keep learning? Decide which cases they still work through by hand, and who walks them through the AI-assisted ones.
- When will you check? Set a date to compare the plan with what actually happened to the hours.
Keep the plan in a shared document in Microsoft 365 or Google Workspace, so remote staff see what their manager sees.
Map the workflow, then reassess
The firms that get the most from AI look at the whole workflow, not one task. HBR's November–December 2026 magazine article, "For Real Performance Gains with AI, Start by Mapping Your Workflows," makes that case using research involving 515 companies. Those companies were startups, not accounting firms.
The research is a field experiment by Hyunjin Kim, Dahyeon Kim and Rembrand Koning, described in an April 7, 2026 write-up by the HBS AI Institute. All 515 high-growth startups received API credits, access to frontier AI models, and technical training. A randomly selected group also received case studies showing how AI-native companies reorganize their workflows, teams, and business models. Compared with the control group, those startups identified 44% more AI use cases, completed 12% more tasks, were 18% more likely to land paying customers, and generated 1.9 times the revenue.
The write-up opens with a line any operations lead will recognize: "A firm is only as fast as the slowest step in its chain of work." The researchers call the challenge the mapping problem.
Startups are not tax practices, and this result does not show that mapping will produce the same gains in your business. It does fit the accounting story. Speeding up preparation moved the constraint to review. Fixing review will move it somewhere else, perhaps to client intake or scheduling. A practical sequence:
- Pick one workflow and list its steps from request to delivery, with an owner for each.
- Mark where AI is used, where review happens, and where work waits.
- Assign the freed hours using the written plan above.
- After one full cycle, find the new slowest step and repeat.
We can help you with that.
NTx AI Consulting works with owner-led and partner-led privately held businesses that are currently using AI tools in everyday workflows, or would like to be. The complimentary consultation is a working session, not a product demo. If a deeper engagement makes sense, Phase 1 is an AI Disruption Readiness Audit. You can learn more at https://ntx.ventures/ai-consulting/.
Sources
- Harvard Business Review article of November 11, 2026. https://hbr.org/2026/11/the-hidden-costs-of-delegating-to-ai
- June 2, 2026 announcement. https://www.current.co/news/crete-professionals-alliance-rebrands-as-current
- Thrive Capital announced in January 2022. https://medium.com/@thrivecapital/welcoming-nitin-nohria-a8c5f3dd22e9
- Forbes reported on June 2, 2026. https://www.forbes.com/sites/annatong/2026/06/02/thrive-holdings-to-bet-1-billion-on-ai-powered-accounting-roll-up/
- article's abstract. https://store.hbr.org/product/the-hidden-costs-of-delegating-to-ai/H09DFY
- Leitgrad's November 12, 2026 summary of the HBR article. https://www.leitgrad.ch/ki-spart-treuhaendern-ein-drittel-der-zeit-entscheidend-ist-wohin-die-stunden-gehen/
- For Real Performance Gains with AI, Start by Mapping Your Workflows. https://hbr.org/2026/11/for-real-performance-gains-with-ai-start-by-mapping-your-workflows
- April 7, 2026 write-up by the HBS AI Institute. https://aiinstitute.hbs.edu/everyone-has-ai-which-firms-are-going-to-win/