AI projects are more likely to deliver value when organisations start with a business outcome, redesign the work around it and keep a person accountable for the result. A Raconteur article featuring PA Consulting’s examples sets out five practical moves, from choosing measurable goals to tracking whether a project is worth continuing.
Watch Desk analysis
What happened
The article describes approaches used by international ground-handling company Çelebi Aviation and law firm Womble Bond Dickinson. Çelebi defines a project’s intended outcome before implementation, and uses AI video analysis to monitor customs warehouses and audit ground-handling operations. Its Comply AI system can flag possible safety or compliance failures for review and training.
Womble Bond Dickinson has required human oversight under its AI policy since 2023. Lawyers must verify AI-assisted work, flag uncertainty and use the firm’s existing supervision or peer-review process. The firm also trains staff and asks lawyers to define what a good result should look like before comparing it with an AI output.
The article’s five recommendations are to start with the desired outcome, redesign workflows where AI can add value, make human accountability explicit, support employee adoption, and measure operational value rather than usage alone. Çelebi, for example, tracks measures such as cycle time, manual effort, quality, compliance and financial impact, and says it revisits a project’s design or scope if results fall short.
Why it matters
These examples make a useful distinction: giving staff access to AI is not the same as improving the work. A tool’s usage numbers or time saved may say little about quality, risk or whether the organisation can do something it could not do before.
The safety and legal examples also put responsibility in practical terms. Potential problems are flagged for people to review, and AI-assisted legal work remains subject to verification and supervision. That is more informative than treating “human in the loop” as a magic phrase.
Our read
The strongest advice is to decide what success means before buying or deploying the tool, then measure whether the change actually delivers it. Otherwise, organisations risk mistaking activity for value, which is a very expensive way to feel busy.
The article presents selected company examples and recommendations, not proof that these approaches will work everywhere. Still, its measures and accountability steps give leaders something concrete to test against their own projects.
What to watch
- Whether organisations define success measures before deployment, rather than after the dashboard arrives.
- How firms assign responsibility for checking AI-assisted work.
- Whether evaluations include quality, compliance and financial impact as well as time saved.
- Whether projects are redesigned or stopped when results do not justify further investment.
Discussion spark: Should organisations measure AI success mainly by time and money saved, or give equal weight to quality, accountability and the work they can now do?
Sources and evidence
- Source update (18 November 2026, 15:33 UTC)
- Source update (18 November 2026, 15:33 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.