A briefing for leaders who are thinking about making AI part of how the organisation works.
At a technology meetup in Germany recently, the conversation over coffee kept coming back to hackathons. Several of the people there worked at organisations that had used Generative AI hackathons to build a culture of innovation, and they had plenty to show for it: people learning quickly, teams showing skills nobody knew they had, working prototypes in days, and colleagues from different functions building together for the first time.
As the discussion went on, a pattern became clear. The hackathon had been a one-off event. The learning was real, and some silos had genuinely come down. The usual working week, however, was largely unchanged. The same handoffs, the same approvals and the same workflows were all still in place, now with AI helping at each step.
That conversation shaped how I think about AI adoption. Copilots deliver real value. They help people draft, analyse and summarise faster, and that gain at the level of individual tasks is an essential first step. The full benefit arrives when the workflow around those tasks is redesigned too. A draft written in ten minutes reaches the people who depend on it faster only when the handoffs and queues around it move at the same pace.
Figure1. Illustrative timeline comparing AI-assisted tasks inside an existing process with long handoffs and waiting, and a redesigned workflow that moves from task to human review to done.
This is where many organisations we speak to are now. The copilot is in place and people are using it well. The next step is redesigning how the work flows, who decides what, and how human judgement is applied where it matters most.
Most workflows were built around the limits of manual work. Handoffs, reviews and batching exist because people have only so many hours in the day and can only hold so much context at once. A weekly report exists because nobody can watch the numbers continuously. When a copilot speeds up the tasks but the surrounding process stays the same, the gains are real but they stay local.
Redesign also changes where people add value. When AI produces the first draft, the initial analysis or the routine routing, human effort moves in two directions. Upstream, people spend more time defining the goal, framing the problem and setting the guardrails. Downstream, they spend more time on judgement: checking quality, handling exceptions and making the final call. Human oversight becomes more central to the work, so roles, team structures and performance measures need to reflect it.
Then there is the human side. Introducing a tool can happen quickly. Changing team structures, incentives and daily habits takes deliberate leadership over months.
Figure 2. Comparison table of the copilot mindset and the redesign mindset across starting question, process, people’s role, decision rights, data, what gets measured and typical result.
Figure 3. Five numbered steps: start with the outcome, remove before you automate, redraw decision rights, redefine roles, get the data ready first.
Map the end goal of a process and ask how you would build it today, starting fresh with AI available. Then compare that picture with how the work actually runs, and look honestly at where it waits, repeats or stalls. Take a monthly performance report. The real outcome is the leadership team knowing which issues need attention this week. Framed that way, the answer might be a timely alert that a person then acts on.
Cut the duplicated steps, manual data transfers and status meetings that only existed to coordinate work done by hand. A useful test for each step is to ask what would go wrong if it disappeared tomorrow. Steps that protect quality, safety or compliance stay, and often become stronger once people have more time to give them.
Agree clearly what AI prepares and where people review and decide. Match the depth of human review to the risk: routine, reversible work may need a light check, while anything external-facing, regulated or high-impact keeps a named person accountable for the decision. Clear lines give teams confidence in the output and keep accountability with people. This is also where regulation is heading. The EU AI Act makes effective human oversight an explicit requirement for high-risk AI systems, so designing it in now is good practice and good preparation.
Move people’s time towards judgement, edge cases, stakeholder relationships, strategic decisions and overseeing AI output. Write these responsibilities into roles explicitly, so oversight is recognised as core work. Then broaden what you measure. Adoption figures show the tool is being used. End-to-end cycle time, quality and the outcomes stakeholders care about show how the whole workflow is performing.
All five depend on the last one: governed, well-defined data. A copilot is only as useful as the data and knowledge it can reach. When three teams define the same customer three different ways, AI repeats the inconsistency. Shared definitions, clear ownership and governed access let a redesigned workflow scale with confidence, and give the people reviewing AI output a reliable basis for their decisions.
Hackathons give people permission to experiment. Redesign needs permission to change how the week runs. That often comes from middle managers, who own the approvals, reports and routines that make up a normal week. Bring them in early. Ask them to map their own workflows, suggest steps that could be simplified, and help set the new decision rights and review points. When managers co-design the change, they become the people who make it last.
It also helps to keep the hackathon energy going in a more lasting form. A small, standing redesign team that takes one workflow at a time, measures the result and shares what it learned turns a one-off event into a habit.
Pick one workflow that matters, map it honestly, and redesign it end to end with AI built in and human review designed in from the start. Choose something visible enough that people will notice the difference, and contained enough that you can deliver in weeks. The copilot gains you already have are the foundation. Redesign is how they reach the whole organisation.
- Published on November 18th 2026.
If you’re wondering where your own organisation stands, our Data and AI Maturity self-assessment is a quick way to find out. Our team helps organisations redesign workflows on a governed data foundation, with human review built in from the start. If you’d like to compare notes on what we’re seeing across the UK and Europe, talk to us contact@clair-x.ai