1st July 2026
The first AI workflows are in, and employees are less than impressed.
Currently there is deserved criticism aimed at businesses using their shiny new AI workflows, which in many cases are falling flat on their face.
The tools sound impressive, individuals feel faster, but the organisation as a whole does not suddenly become dramatically more productive. The return on investment is often harder to prove than expected.
That does not mean AI workflows are a dead end. It means most companies are still on v1.
I read a great article recently making the comparison to when factories were first electrified, they did not immediately become modern factories. They replaced steam-powered machines with electric machines, but kept the same layout, processes, and assumptions. The technology was new, but the operating model was old.
AI is going through the same transition.
The first stage is the lightbulb stage. New tools help individuals do existing tasks a little faster. People write emails more quickly, summarise documents, generate code, create first drafts, or analyse information. This is useful, but it does not fundamentally change how the business works.
The second stage is where many companies are heading now. AI starts to handle larger parts of the workflows. It can screen candidates, process support tickets, investigate changes in stats, or move information between systems. These workflows are more automated, but they are still usually plugged into existing company structures.
If AI helps individuals and teams produce more output, but decisions still go through the same human approval layers, the bottleneck just moves. Work gets faster, but the organisation does not. More things wait for review, approval, prioritisation, or sign-off.
I'm imagining a v3 is where the big change happens. This is when companies stop asking, “How can AI help this task?” and start asking, “How should this process work if it is designed to be done by AI from the start?”
At that point, AI will not just prepare work for humans. It will increasingly interpret signals, route decisions, trigger actions, and connect workflows without waiting for people at every step. Human oversight will still matter, especially for quality, risk, ethics, and direction. But humans will not necessarily be inside every loop.
That is where the productivity gains really start to kick in. Not because AI writes a faster email, but because the entire cycle time of a business process changes. Processes that used to wait for someone to notice a signal will start to happen automatically.
This won't happen evenly, and won't be simple. Some companies will automate poor processes and make them worse. Some will remove human judgement where it is still needed.
But the direction is clear. AI workflows v1 do feel clunky. v2 will be better and cut out a lot of manual work, but v3 might remove vast swathes of human interaction.
Whether that is a good thing for the economy as a whole remains to be seen.
The important point is that today’s AI workflows should not be judged as the finished product. They are the first iteration in a much deeper reorganisation. The real impact will come when companies stop adding AI to old processes and start rebuilding processes around what AI makes possible.