Implementation is crucial to whether AI compounds a team’s strengths or automates bad habits
A seed dropped into fertile soil takes root and grows fast. The same seed dropped into poor, compacted soil either struggles or fails outright, and the seed itself never changes. Guy Fenton, global director for Akai at Deel, describes artificial intelligence in almost identical terms: the technology does not decide what a team becomes, the conditions it lands in do.
"The same tool dropped on two different teams produces opposite results," said Guy Fenton, Akai Global Director at Deel. "AI is ultimately the amplifier. If you drop it on a very good, high-trust team, it obviously compounds. But if you drop it on a relatively dysfunctional one, you start to automate the dysfunction even faster and even more so."
For Fenton, that single observation reframes the whole adoption conversation. AI is not an upgrade a business installs once and moves on from. It is a magnifier held over whatever already exists in a team, for better or worse.
What good implementation actually looks like
Having been at Deel since 2020, Fenton has watched enough rollouts, inside Deel and with its customers, to identify what separates the ones that stick from the ones that stall. It starts with locating the actual pain, not the theoretical one.
"You really have to understand the areas of the business that are falling behind, in terms of what are the manual, repetitive workflows or what's taking the most time for employees," said Fenton. "You need people on the other side to implement it and to actually understand the pain points that your employees are going through, in order to generate more interest."
Once that groundwork is done, Fenton said the second piece is showing staff what the finished state looks like before asking them to trust it, rather than dropping a new tool on their desk and hoping curiosity does the rest. That means giving employees a look at the output an AI-assisted workflow can produce, then supporting them through the change instead of leaving them to work it out alone. Fenton also pointed to a discipline that gets skipped once the excitement of a launch fades: reviewing whether the numbers are still moving, not just checking them once and declaring success.
Deel built its own answer
Deel's own experience is the case study Fenton keeps returning to, largely because no outside vendor could handle the scale it needed. The business transacts billions of dollars a month, holds thousands of bank accounts across hundreds of countries, and decided roughly 18 months ago to build automation internally rather than wait for the market to catch up.
It started with the team carrying the heaviest manual load.
"We wanted to start with the payment operations team initially, because that's where a lot of the manual processes lie," said Fenton. "We started off with that, and then obviously it worked so well, we started expanding it across many different teams thereafter."
That expansion now touches finance, tax, treasury, benefits, HR, customer support and go-to-market functions. According to Deel's own published figures, the platform, now offered externally as Akai, handles more than 250,000 cases and saves in excess of +1 million hours to date across the business, with reconciliation work that once took more than 20 days now finished in minutes. The tool has been deployed to all roughly 7,000 people at Deel, spanning legal, marketing, sales and finance as well as operations.
The real return isn't hours
But Fenton says the real metric is not about hours saved.
"It's actually what your best people do with the hours that you give them back," said Fenton. "At Deel, a lot of the work our teams now move towards is customer-facing work, or handling complex problems, or building relationships with clients."
That shift changes what strong performance looks like in practice, and Fenton is blunt about the scale of it.
"A very strong performer with AI does the work of three people or four people, faster, with better judgement," said Fenton. "And that changes the maths on team size and team shape."
Fenton was careful to frame this as a capability story rather than a cost-cutting one: the point is not fewer people doing the same work, but the same people doing markedly more of it.
Closing the gap is HR's job
That same capability shift has a harder edge, and it is one Fenton thinks Australian HR teams are not yet talking about enough. He spends his days across professional services clients hearing plenty about efficiency targets and very little about what happens to the people delivering them.
"AI is amazing when it comes to efficiency gains and productivity gains, yes, absolutely," said Fenton. "But then that has a roll-on effect to people management. You can actually start spending a lot more time with your people, and curating a lot more people skills."
Left unmanaged, that shift widens the distance between a team's strongest and weakest performers rather than lifting everyone together, and Fenton frames the response as squarely an HR responsibility.
"It only becomes a cliff if you don't invest in carrying people across it," said Fenton. "You have to start enabling people very early on that might not be accustomed to using AI tools, and that reskilling job is HR's to own."
The job gets more human
The part of the shift Fenton did not expect is how little of it ends up being about machines at all.
"The surprise for me of running an AI-native team is that the job actually becomes more human, not less," said Fenton. "When the machine starts doing all the grunt work, what's left is judgement, what's left is taste, what's left is relationships."
See how Akai works here.
This article was produced in partnership with Deel
