Most AI implementations go straight to speed. What they build rarely lasts.
Efficiency was the goal. That part was right.
In our case, that person was a colleague who brings more than task execution to her role. Judgement. Experience. Context that doesn’t show up in a brief. Automating her tasks wouldn’t improve any of that. It risked removing the very things that made the output good.
More importantly, I hadn’t asked what she wanted. What was actually hard. What she’d change if she could.
Proposal complete before business lead reviews Dovetail Digital internal.
Down from 3–4 hrs for brief to first draft Dovetail Digital internal.
Of companies see zero productivity gains from AI NBER, ~6,000 executives, Feb 2026
That last figure matters. A 2026 NBER study of approximately 6,000 executives found that 90% of companies report zero measurable productivity impact from AI, despite 69% already using it. The gap isn’t the technology. It’s how it’s being introduced. (NBER via Fortune, February 2026)
| Tool-forward (most common) | Capability-led (what works) |
|---|---|
| What can we automate? | What does this work actually involve? |
| Faster inconsistency | Repeatable, defensible output |
| Governance arrives late | Governance is light and built in from the start |
| Scales poorly — dependent on individuals | Scales because the process carries the logic |
1. What does the work actually involve? Not the task list, the thinking, the judgement calls, the parts that can’t be templated.
2. What’s creating friction, and for whom? The person closest to the work usually knows where the drag is. The person running the business often guesses from the outside.
3. Where must human judgement stay? This isn’t a caution about AI, it’s a clarity question about accountability. Who checks the output? Who owns the decision?
We work primarily with Australian service businesses, professional services firms, hospitality operators, member organisations, and clubs. Across those environments, the same dynamic plays out.
A general manager who wants to cut reporting time, but whose team is stretched. A senior manager who can see the efficiency gains but isn’t sure what’s appropriate to share with the board. A business lead who wants AI-assisted client communications, but where relationship context is everything.
Most start with tasks rather than understanding the work. When AI is introduced before processes are clear and accountability is defined, organisations get faster inconsistency rather than genuine capability. NBER research (2026) found 90% of companies report zero measurable productivity impact from AI despite widespread adoption.
Human at the helm means humans set the direction, define quality, and remain accountable for outcomes before and after AI is involved. It differs from ‘human in the loop,’ which typically means passive review of AI output at the end of a process.
INGRAIN is a structured AI capability framework that sequences adoption across four stages: Get Clear, Get Fluent, Get Consistent, and Get Growing. It prioritises governance, shared practice, and leadership alignment before scaling tools. Dovetail Digital is a Certified INGRAIN Implementer.
Before introducing any AI tool, service businesses should understand what the work actually involves beyond the task list, identify where friction sits from the perspective of the people doing the work, and define where human judgement must stay. These three questions determine which tools and workflows are worth building.
Still feeling stuck? You’re not alone but you don’t have to figure it out solo.
Our DMA helps you cut through the noise and focus on what matters most.
