By David Sharpe | Managed Services Lead, Mantel
Key takeaways for business leaders
- Most enterprise operations still run on a model designed for human-paced work.
- AI benefit will not scale inside a managed services contract that still rewards activity – the next model must price outcomes, share automation – gains, and keep humans accountable for judgment.
- At a top Australian mortgage aggregator, an agent now triages every incident in under a minute, for less than a dollar per event.
- Four questions will show whether your operating model is ready for agentic operations. Most leaders can confidently answer two or three.
Adapting to Agentic Operations
Managed Service Providers (MSPs) have frequently adjusted the operating context, governance approach and commercial model to keep pace with technology.
Each shift was driven by the logic that when the work changes, the way you measure and pay for it has to change as well. Today we see the next shift happening with MSPs adapting to meet the opportunity of agentic operations.
With the advent of agentic operations:
- Adversaries can now operate at machine speed
- Customers increasingly expect an always-on, personalised service
- Businesses want measurable AI ROI.
Previous Managed Service models, built around inputs like people, hours, and SLAs rather than the outcomes the business actually tracks, are inherently misaligned. Once the model is built around outcomes rather than hours, the economics shift quickly.
That economic shift is already visible in production. At a top-three Australian mortgage aggregator we run managed services for, an agent now triages every incident, cutting time-to-diagnosis from 30-60 minutes to under one minute at less than a dollar per event. Our team stays in the loop, applying the expert judgement that matters, but the overall time saved is dramatic. The workflow was chosen on its merits – it’s relatively safe to get wrong, costs almost nothing to run, and is a space where consistent agent execution pays back beyond the human time saved, with every alert getting the same structured first response that arms the responder.
The change is in what operations teams own, how they are organised, and what they are accountable for. Agents take the busy work; people own the judgment, the design and the escalation calls that matter.
Across our engagements, four questions consistently separate the initiatives that keep paying back from the ones that stall. They are worth putting to your own operations approach.
“We can no longer afford to pay good people to do busy work. The operating model has to evolve as fast as the work itself.”
David SharpeManaged Services Lead, Mantel
Four questions worth putting to your own operations approach
1. Are you paying for activity, or for outcomes?
Where activity is the performance metric, activity tends to sustain or quietly grow. The work the operations team does should map back to measurable business value and not activity.
2. Where does the boundary between you and your MSP actually sit?
The previous MSP models often ran on hard boundaries – cloud, ITSM and data platforms inside the supplier’s perimeter, with the client kept at arm’s length from both the work and the tools. That model breaks under agentic operations. Agents need current context from the client’s environment, and the client needs visibility into how the agents acting on their estate are behaving – which actions they took, why, and where a human stayed in the loop.
The model that scales is shared accountability, with the MSP’s approach personalised to each client’s platforms, data and ways of working rather than imposed from a generic playbook. The engagements that stall are the ones where either side still expects a black box.
3. Where do you start?
The organisations making the most progress on agentic operations begin by understanding the work being done today. Just as captured, structured, high-quality data is what unlocks business insight, captured, structured, high-quality knowledge of how IT actually runs is what unlocks agentic operations. With that foundation in place, you can reliably score workflows, work out the real ROI of each, and pick the right first candidates for an agent.
4. Does your model reward automation, or punish it?
A contract built for agentic operations puts both sides on the same side of the table – outcomes are priced, gains are shared, and both parties are incentivised to make the work itself shrink.
We have identified three contractual approaches that often work:
- Outcome-priced – a defined set of outcomes at a defined price, with the supplier free to choose the human/agent mix.
- Risk and reward shared – automation gains are tracked transparently and split between supplier and customer.
- Joint capability uplift – supplier commits to improving the environment; the customer commits to sponsoring the change.
If you can confidently answer all four, you are already further along than most. If you cannot, those are the conversations Mantel is having with operations leaders right now.
Mantel is one of the few Australian consultancies with deep production experience taking AI agents to scale from agentic commerce at a tier-one supermarket, to claims automation at most major Australian insurers, to Australia’s first agentic home loan experience. The same discipline that turns raw data into business insight – captured, structured, high-quality – is what turns operational knowledge into agentic operations.
See how we’re helping businesses scale with AI-first solutions
