AI-first company

Process redesign · Redesign

Redesign the workflow instead of adding a tool to it

Giving people an AI tool inside an unchanged process changes little. Decide again who does each step, in what order, and where a person checks.

By
Last reviewed
Corrections
Report an error
When Do next. Once your first uses of AI are running.
Effort High (months, several teams)
Evidence Moderate

One large survey links workflow redesign to profit from AI. It relies on what companies report about themselves, and comes from a consultancy. One randomised trial in breast screening compared a redesigned process with the old one and found more cancers with less work. No study in ordinary business has compared a redesigned process with the same process plus a tool.

Fits
Industries All industries
In the guide Mid-market companies

What it is

Go through the process step by step and decide again: which steps AI does, which people do, what order the steps come in, and where output is checked. That is redesign. Adding a tool means the same people do the same steps, with an AI assistant on the side.

What the evidence shows

  • Redesign sets apart the companies that profit. In McKinsey’s 2025 survey of 1,993 respondents, about 6% of organizations credited AI with 5% or more of operating profit. Those organizations were nearly three times as likely as others to have redesigned specific workflows from the ground up. Of 31 practices the survey looked at, this was one of the practices that set the profitable firms apart most clearly.
  • Research on tasks points the same way. A study that scored 18,156 work tasks concluded that getting the benefit of machine learning usually requires redesigning what a job contains.
  • Just handing out the tool changes little. Across the Danish labour market, chatbots were widely adopted without any change in earnings or hours. What did change was how tasks were organized. In a field experiment with 7,137 workers at 66 firms, workers randomly given an office AI assistant spent about two hours less on email a week, and the amount and mix of their other work did not change. The figures come from the software’s own activity records. Three of the four authors worked for Microsoft, which sells the assistant.
  • One redesign has been tested in a randomised trial. In a Swedish breast-screening trial with 105,934 women, AI decided which scans went to one radiologist and which to two, and marked suspicious areas. Against standard reading by two radiologists, 29% more cancers were found per 1,000 women, false positives did not rise, and reading work fell by 44.2%. The setting is a regulated health service with a fixed protocol, not ordinary office work.

How to do it

  1. Choose one process where an AI pilot has already shown a gain on a single step.
  2. Map the whole process from first input to final output, including waiting time and hand-offs.
  3. For each step, decide: AI alone, person alone, or both. Test this where you can.
  4. Remove steps that existed only because the old way was slow.
  5. Define where a person must check AI output, and what the checker looks for.
  6. Name one owner for the whole process.
  7. Run the new process next to the old one for a short time, and compare.

What to measure

  • Time from start to finish of the whole process, not per step.
  • Quality of the final output.
  • Number of hand-offs.

Where it fails

  • Redesigning before you know what AI can do in your setting. Run a small pilot first.
  • Removing every check. The same survey found that the best performers were more likely to define when a person must check AI output.
  • Calling integration redesign. In surveys, “built AI into our processes” is not the same as redesigning them. A Federal Reserve analysis found that US small firms that had built AI into their processes had no better hiring expectations than firms still experimenting. In payroll data from Gusto, which sells payroll software, light and deep users showed the same hiring gain.
  • Ignoring the people who do the work. The people who do the work know why the odd steps exist. See protect psychological safety.

By company size

  • Micro and small: few old systems and short decision chains mean you can redesign a process in weeks.
  • Enterprise: processes cross departments and systems. Give the redesign an owner with authority over the whole chain.

Company cases

Companies that did this

Sources

  1. McKinsey & Company (2025). The State of AI: Global Survey 2025. McKinsey Global Survey, published 5 November 2025. What this study did and found
  2. Brynjolfsson, Mitchell and Rock (2018). What Can Machines Learn and What Does It Mean for Occupations and the Economy?. AEA Papers and Proceedings, 108, 43–47. What this study did and found
  3. Humlum and Vestergaard (2025). Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI. NBER Working Paper 33777 (revised March 2026). What this study did and found
  4. Dillon, Jaffe, Immorlica and Stanton (2025). Shifting Work Patterns with Generative AI. NBER Working Paper 33795; arXiv 2504.11436, version 4 (November 2025). What this study did and found
  5. Hernström, Josefsson, Sartor, Schmidt, Larsson, Hofvind, Andersson, Rosso, Hagberg and Lång (2025). Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel-group, non-inferiority, single-blinded, screening accuracy study. The Lancet Digital Health, 7(3), e175–e183. What this study did and found
  6. Tremper and Bowen (Gusto) (2026). Small Businesses that Adopted AI Are Hiring Faster. Gusto Insights, September 2026. What this study did and found
  7. Aarons and Sarkar (October 2026). Federal Reserve Bank of New York, Liberty Street Economics. Analysis of AI use and hiring expectations in the Small Business Credit Survey. libertystreeteconomics.newyorkfed.org

Claims on this topic, checked against these studies: