Understand what AI could be worth to your business
AI can change how much work your team can handle, how well they do it and what your business can offer. The useful question is where those improvements could matter to you, and what it would take to achieve them.
We help you assess the opportunities, examine the assumptions and agree how you will judge progress. As you put AI to work, your own evidence gives you a stronger basis for deciding what to continue, change or expand.
Independent research is clear on two things: used well on the right work, AI makes people markedly faster and better; used without guidance, it can make them worse. We help you get the first and avoid the second.
Start a conversation
What the research shows
The most rigorous studies of AI at work point the same way. The gains on the right tasks are large, less experienced people gain most, and the organisations that encourage and train their people get far more from it.
Faster, better work
Noy and Zhang (2023); Dell'Acqua et al. (2023)
Less experienced staff gain most
Brynjolfsson, Li and Raymond (2025); Dell'Acqua et al. (2023)
Used badly, AI makes people worse
Dell'Acqua et al. (2023)
Encouragement multiplies the benefit
greater benefits where employers actively encourage use; training from the employer raised adoption
Humlum and Vestergaard (2025)
On a task outside what AI does well, consultants using it were 19 percentage points less likely to reach the right answer than colleagues working without it. Knowing where AI helps, and where it does not, is part of the value.
What would better look like?
Start with the work you want to improve. A useful business case connects AI to an outcome people can recognise and measure.
More capacity
Reducing repetitive work could give your team more time for customers, complex tasks or growth. The question is what they would do with that capacity and whether the demand is there to use it.
In a study of 25,000 workers, time saved averaged 2.8% of working hours, rising to 6.8% where employers encouraged use (Humlum and Vestergaard, 2025).
Better quality and service
AI may help people prepare more thoroughly, work more consistently or respond sooner. The measures might be turnaround time, rework, accuracy or customer experience, with appropriate human review built into the process.
In controlled studies, the quality of professional writing rose 18% and of consulting work by more than 40% (Noy and Zhang, 2023; Dell'Acqua et al., 2023).
Lower costs
Some changes may reduce expenditure or avoid an additional cost. A credible estimate identifies which costs would actually change and includes what it takes to introduce, support and run the new approach.
Better control of risk
Clearer policies, visibility and review can help a business manage how AI is used. Progress might mean fewer policy breaches or more consistent checks, rather than a speculative financial saving.
Used outside its strengths, AI made consultants 19 percentage points less likely to be right (Dell'Acqua et al., 2023). Clear rules and guidance protect quality as well as data.
New capabilities
AI may make a service, analysis or way of working feasible that was previously too time-consuming or expensive. Small experiments help test whether the opportunity is useful and whether customers or colleagues would use it.
Newer customer support staff resolved 34% more issues per hour with an AI assistant, as it spread the practice of the best performers (Brynjolfsson, Li and Raymond, 2025).
Time saved needs somewhere to go
An hour saved is an opportunity. Its value depends on what happens next.
If your team uses that hour to serve another customer, improve a decision or clear a backlog, the business may gain something tangible. If it simply disappears into the working day, the financial benefit may be limited.
We distinguish capacity released from expenditure reduced and additional income earned. That helps you see what a potential saving means in practice and what needs to happen for it to become value.
The research bears this out. In Denmark, most workers using AI said they spent the time saved on other work, and the study found no measurable change in earnings or hours worked (Humlum and Vestergaard, 2025). In a UK government department, staff saved time on written tasks, but the evaluation found no evidence that this improved overall productivity (Department for Business and Trade, 2025). Time saved only becomes value when someone decides what it is for.
A business with 50 people. At the cautious 2.8% of working hours, AI would release about 2,300 hours a year, the capacity of between one and two people. At the 6.8% seen where employers encourage use, it would be about 5,600 hours, more than three people. Whether that becomes value depends on what those hours are used for.
Illustration based on 50 people working about 1,650 hours a year and the figures in Humlum and Vestergaard (2025). Not a forecast for any business.
Build the case, then test it
Establish the starting point
Understand how the work happens today: who does it, how often, how long it takes and where delays, costs or quality problems arise. Use your own information where it is available, and identify gaps that need further investigation.
Make the assumptions visible
Estimate the difference AI could make, allowing for adoption, checking outputs, training, implementation and ongoing support. Separate measured information from estimates and explore how the result changes under different assumptions.
Published research and sector benchmarks can inform this work. They provide context, while your circumstances determine how well the findings apply.
Learn from use
Agree measures before starting, then review what happens in practice. Look at use, quality and business outcomes together. That evidence helps you decide whether to adapt the approach, extend it or stop.
Why results vary
Two UK government studies of the same tool reached different conclusions. Across 20,000 civil servants, Copilot users reported saving 26 minutes a day (Government Digital Service, 2025). In one department, the evaluation found time savings but no clear improvement in productivity (Department for Business and Trade, 2025). The tool was the same. What differed was how it was used.
saved by Copilot users, self-reported from ranges given in the survey. The report notes Copilot needed human oversight on nuanced or complex material.
Government Digital Service (2025)
that time saved on written tasks improved overall productivity. Some tasks took longer where outputs were poor.
Department for Business and Trade (2025)
The right work
AI is strong at some tasks and misleading on others that look similar. Start where the evidence is strongest: drafting, summarising, analysis and routine communication.
People who know how
Benefits are markedly greater where employers encourage use and provide training. Access alone is not enough.
Rules and review
Clear policies, visibility of how AI is used and regular review keep quality up and risk down while the benefits build.
Keep value under review
The first business case is a starting point. People's confidence, working practices and the available technology will change as you learn.
Requisite Cadence, our ongoing AI leadership and governance service, brings usage, adoption, training, risks and opportunities into regular review. Each month, the Intelligence Statement includes the value AI is delivering, measured against starting points agreed with you. Each quarter, your leadership team uses that evidence to decide what to continue, change or expand.
See how we help →Explore the evidence for your sector
Our research examines the assumptions behind AI adoption and potential returns. For law firms, we offer a complimentary AI value briefing based on public information and sector evidence, with illustrative scenarios to support an informed conversation.
Explore our reports and research →What would make AI worthwhile for you?
Tell us about the work you want to improve or the opportunity you want to explore. We can help you identify what to investigate and how to assess whether it is worth pursuing.
Start a conversationSources
- Brynjolfsson, E., Li, D. and Raymond, L.R. (2025) 'Generative AI at work', The Quarterly Journal of Economics, 140(2), pp. 889–942. Available at: https://doi.org/10.1093/qje/qjae044 (Accessed: 14 October 2026).
- Dell'Acqua, F., McFowland III, E., Mollick, E.R., Lifshitz-Assaf, H., Kellogg, K.C., Rajendran, S., Krayer, L., Candelon, F. and Lakhani, K.R. (2023) Navigating the jagged technological frontier: field experimental evidence of the effects of AI on knowledge worker productivity and quality. Harvard Business School Technology & Operations Management Unit Working Paper No. 24-013. Boston, MA: Harvard Business School. Available at: https://ssrn.com/abstract=4573321 (Accessed: 14 October 2026).
- Department for Business and Trade (2025) Evaluation of the M365 Copilot pilot in the Department for Business and Trade. London: Department for Business and Trade. Published 28 August 2025. Available at: https://www.gov.uk/government/publications/microsoft-365-copilot-pilot-dbt-evaluation-report (Accessed: 14 October 2026).
- Government Digital Service (2025) Microsoft 365 Copilot experiment: cross-government findings report. London: Government Digital Service. Published 2 June 2025. Available at: https://www.gov.uk/government/publications/microsoft-365-copilot-experiment-cross-government-findings-report (Accessed: 14 October 2026).
- Humlum, A. and Vestergaard, E. (2025) Large language models, small labor market effects. Becker Friedman Institute Working Paper No. 2025-56. Chicago, IL: University of Chicago. Available at: https://bfi.uchicago.edu/wp-content/uploads/2025/04/BFI_WP_2025-56-3.pdf (Accessed: 14 October 2026).
- Noy, S. and Zhang, W. (2023) 'Experimental evidence on the productivity effects of generative artificial intelligence', Science, 381(6654), pp. 187–192. Available at: https://doi.org/10.1126/science.adh2586 (Accessed: 14 October 2026).