|
AI, Applied
AI is moving from helping people do work to doing parts of the work itself.
AI, Applied
What changed. What matters. What to do about it.
Issue 01 · Fortnightly · 5-minute read
|
|
|
AI is moving from helping people do work to doing parts of the work itself.
For the last couple of years, AI has mostly sat inside a chat window: you asked, it answered, and you decided what happened next. That's still true for a lot of AI use today.
|
|
2024–25
“Help me do this.”
|
→ |
2026
“Do this piece of work.”
|
|
|
|
Over the past 6–12 months, that's started to change: AI can now carry out genuine pieces of a workflow with limited supervision — gathering information, checking it, handling routine cases — rather than needing direction at every step. Independent researchers who track this have found a well-established trend of AI's reliable task length roughly doubling every seven months since 2019, and some evidence — still contested by the researchers themselves — that the pace may have picked up further recently. Westpac is a concrete local example: its AI agents now extract and verify payslip and bank-statement data on mortgage and credit-card applications, saving an estimated 150,000 banker hours a year — while the bank is explicit that its bankers, not the AI, still make the actual lending decisions.
The clearest lesson on how far to trust it comes from Klarna. Its AI took on two-thirds of customer service tickets and looked, for a while, like a wholesale replacement for its support team — until the company had to walk back some of the staff cuts it made, because the AI struggled with complex, ambiguous or emotionally charged cases. It now runs AI and people together deliberately, not AI instead of people.
So the shift isn't “AI can run your business.” It's that there are now more genuine, provable pieces of everyday work you can hand to AI — provided a person still owns the outcome.
|
| What This Means For Your Business |
|
|
Customer Service
|
Handling routine, well-defined enquiries start to finish — with complex or emotionally charged cases still going to a person, not the AI.
|
|
|
Finance
|
Reconciling transactions and flagging discrepancies for someone to review — not replacing the judgement calls entirely.
|
|
|
Operations
|
Monitoring workflows continuously and surfacing exceptions — the ongoing checking, not the decisions about what to do next.
|
|
|
Sales
|
Researching prospects and drafting outreach, so a salesperson starts from a warm brief, not a blank page.
|
|
|
HR
|
Handling the administrative steps of onboarding, leave requests and policy questions — while people decisions stay with people.
|
|
|
|
|
|
Three actions, not a checklist
|
1
|
Find repetitive knowledge work.
Look for work where people repeatedly gather information, make predictable decisions, or move information between systems.
|
|
2
|
Pick one workflow — not an “AI project.”
Understand its current time, cost and quality first, then test whether AI can materially improve it before committing further.
|
|
3
|
Define the boundaries — and keep checking them.
Decide what AI can do independently, what needs a person to check, and what stays fully human. Klarna's experience is a reminder to revisit that line as you go, not set it once.
|
|
|
|
|
|
Many businesses have an AI access-control gap they may not know about
IBM's 2026 global study of 602 organisations found AI-related data breaches cost about $6M on average, versus $4.99M overall — and among the roughly one in five organisations that had an AI-related security incident, 92% were missing basic controls like role-based access and multi-factor authentication on the AI system involved.
Source: IBM →
|
|
AI vendors are moving closer to the actual work of implementation
Microsoft ($2.5B) and AWS ($1B) have each launched billion-dollar-scale initiatives to embed their own engineers inside client businesses. It's a sign that redesigning workflows and managing the change around AI — not buying the software — is where much of the value is actually won or lost.
Source: CNBC →
|
|
Most businesses using AI can't yet show it's paying off
A National Bureau of Economic Research survey of nearly 6,000 executives across the US, UK, Germany and Australia found 69% of firms use AI — but nine in ten reported no detectable impact on employment or productivity over the past three years.
Source: NBER →
|
|
|
One Question For Your Business
What repetitive work is taking up your team's time that AI could now help with?
|
|
|
That's it for this edition — short on purpose. If one of these changes what you do this week, it did its job.
|
|
Mia Pursehouse · Business Transformation Consultant
|
|
|
AI, Applied is an executive briefing for leaders who don't have time to follow AI but need to know what it means for their business.
|
|
AI transparency: AI tools assisted with research and drafting. Final selection, analysis and editorial review by Mia Pursehouse.
|
|
Read past issues
·
Manage preferences
·
Unsubscribe
Mia Pursehouse, Buttaba, NSW 2283
Sent to Reader
|
|
|
|
