AI Saves Time, but Does the Organisation Ever See It?

Most weeks, this piece is built around one new conversation. At aibl, we’ve spent the past few months talking to operators about how they’re using AI inside their organisations. Nineteen interviews in, one question kept coming up: people were finding ways to save time, but were their organisations seeing the benefit?

We went back through those conversations. Six came at the question from different parts of the workflow. Together, they point to the same distinction: AI can make a person faster without making the organisation faster.

Amin Mrini, Group Chief Commercial AI Officer at Informa, gives a simple example. Generative AI might produce a presentation much faster, but if the decision it supports still waits for the same monthly meeting, the wider process hasn’t accelerated.

“You haven’t saved time,” Amin says. “You’ve just given people more time to over-polish an asset.”

He’s measuring time at the level of the whole workflow. The presentation is ready sooner; the decision it exists to support is not.

Dr Laura Weis, Global Human AI Strategy Lead at WPP, sees the same divide. “On an individual level, a lot of people are feeling that efficiency, but we can’t scale it,” she says.

What happens when the faster output meets the rest of the organisation?

Is the saving real?

Laura describes “shadow work”, where AI produces an apparently fast first pass but weaker output creates correction and editing downstream. “Usually your strong performers get tied into a lot of correction and editing, and that’s not where they’re best placed,” she says.

In that case, time has not been eliminated so much as transferred. The proper test is whether the end-to-end task becomes faster while quality holds steady.

That can happen locally when the person saving the time also controls what happens next. aibl co-founder Terry O’Dwyer uses Notion, Make.com, and an OpenAI Assistant to turn voice notes into actions, commercial flags, and a searchable record. Work that used to take 20 to 30 minutes, and often didn’t happen, now requires seconds of input.

He still reviews every output and decides what to act on, so there is a downstream step. It’s simply under his control rather than distributed across teams. Removing the friction means the work gets done. Over time, the structured record could also turn knowledge trapped in one founder’s head into something the business can use.

What happens at the handover?

Things change when the output crosses into somebody else’s work. Sometimes the next stage isn’t constrained by a fixed meeting or approval cycle. The problem is simply that the people receiving the output already have more than they can act on.

Ben Lee, Head of Data and AI at Bidwells, saw this when the firm tested call-analysis products. The tools surfaced useful insight, but their value was limited by whether the team had the capacity to act on it.

“These tools are amazing,” Ben says, “but they’re only as good as the humans on the other side making the decisions.”

Ben’s example doesn’t tell us how many hours, if any, were saved. What it does show is a broader constraint: AI can increase the volume or speed of useful output without increasing the organisation’s ability to act on it. If the team is already overloaded, the insight may simply sit unused.

The handover itself isn’t necessarily the problem. At Just Move In, the company’s assistant, Jay, handles the opening digital conversation in its first live vertical, energy, and builds a customer profile and dynamic checklist that pass to a human specialist, so nobody starts again. Ross expects specialists to manage more customers through shorter interactions. “They’ll manage more customers in a day, but actually the interactions will just be a bit more condensed,” he says. That’s the intended gain, not yet a measured productivity result.

Early indicators are positive: appointments booked by Jay convert at similar rates to the human team’s, while Trustpilot remains at 4.9. Every conversation is also reviewed by AI and humans, with corrections fed back into the system. The difference is that context carries across the handover, with a review loop to catch mistakes.

Measure the outcome

Amin looks at the same question at organisational level. Chasing separate use cases, in his view, can leave organisations with a collection of tools without shifting overall productivity. He looks instead at end-to-end workflows and measures tied to the wider system: sales team throughput, cost of conversion, pipeline growth.

Inside Make, Sara Maldon, who built its AI adoption programme, Miriwa, takes a harder line. She doesn’t count time saved as a success metric on its own. If a project saves time but doesn’t shift a core business number, it doesn’t count. Make tracks measures including revenue, go-to-market speed, time-to-fill in hiring, and revenue per full-time employee. Those measures improved in the areas where Miriwa invested most, although Sara doesn’t claim direct cause and effect.

Terry’s example and Sara’s rule are measuring value at different levels. Terry can point to a task that’s dramatically faster and more likely to happen. Sara is asking whether an organisational AI project moves a business outcome. Hours saved can tell you something useful. On their own, they can’t tell you whether the wider organisation has benefited.

What can the organisation do now?

AI’s first effect is often speed. The prize is what the organisation does with the space it creates.

That space can be used to serve more customers, make better decisions, reduce strain or build new products. But the workflow has to be redesigned around its real constraint, and the output has to remain usable as it moves between people. The result then has to show up in the measure the organisation set out to change. Sometimes a local outcome is enough, as Terry shows. At organisational scale, Amin makes a sharper distinction. A cost saving is banked once; growth compounds.

After asking how much time AI saved, ask what the organisation can now do that it couldn’t do before.