AI ROI

target readers-cv

By Matt Knibbs

AI ROI depends less on the technology itself and more on ensuring the operating model is effectively augmented with it.

Matt Knibbs is an experienced organisation design practitioner working in public, private and charitable sectors across the UK and internationally. As AI adoption accelerates, many organisations are investing in point solutions without fully realising the intended benefits. Organisations will achieve lasting AI ROI when they consider the end-to-end operating model rather than automating isolated tasks or processes. By designing relevant feedback loops and continuously assessing their readiness for AI, they can adapt to rapid technological change and create lasting value.

Designing the operating model around AI, not the other way around

An AI chatbot was purchased to replace most of the customer service team with promises of efficiency, consistency, and 24/7 responses for customers.

The challenging cases arrived. The point solution only addressed a proscribed set of rules and scenarios. The actual work that customer service teams did didn’t feature in any chatbot setting or workflow. Customer dissatisfaction grew, and the few people left fielding calls ended up dealing with more volume, resulting in re-hiring more humans.

Forrester’s Predictions 2026 report highlighted that 55% of employers regretted AI-attributed redundancies, predicting that around half of these would be reversed. Gartner forecasts that by 2027, half of the organisations that reduced customer-service staff for AI will rehire humans to perform similar roles. The ‘so what’ is not that AI technology is bad, in fact with customer service it is helping to manage higher volumes of work, but labelling it as a way to remove personnel cost can create more risk and not bring about the savings originally anticipated.

It’s interesting that a technology that we know is faster, able to assimilate more data, and better support decision making, often fails to deliver expected results.

How does the operating model shape the value AI creates?

An AI ROI may originate from cost reduction and efficiencies, revenue growth through quicker selling, improving decision quality, or using much more available data to assess risk.

It is not in dispute what can be achieved with good AI implementation, and successful examples of dramatic compression of process time exist. The more important question for leaders is not the time saved, but what parts of the organisation must be redesigned for that saving to turn into lasting value.

Companies fail or succeed because of two things:

  • Use cases, often based on a leadership desire to quickly save cost, without clear organisational outcomes or understanding of current baselines.
  • Tools are deployed before the organisation is ready, automating dysfunction and speeding up a mess, rather than improving it.

Before adopting AI, understand how work, information and decisions flow through the organisation and the associated governance mechanisms. These will lead you to make the right choices on AI utilisation in the context of your operating model and how that will positively impact organisational outcomes.

How can you consider AI within the end-to-end system?

Consider a multinational organisation wanting to deploy a visualisation tool to track organisational health. The technology solution is good, but if local data, reporting practices, and information flows differ between countries, it results in an incomplete picture. The constraint is not the AI or the visualisation tool itself, it is the absence of an end-to-end view on how data is created, stored, moved, and interpreted.

To maximise ROI from AI, organisations must first codify how value is created for their customers. It starts with understanding what activity happens, the decisions that get made, the systems and applications that support this, and how everything is organised (the teams, departments, functions that make up the organisation). It is not simply looking at job descriptions as a significant amount of work happens in organisations that is never written down, fully understood, or even agreed upon.

A thorough review helps make clear which parts could be automated, how that impacts other areas, the dependencies, and ways in which other benefits can be realised. Rather than focusing only on cost control, use AI to augment human work, unleashing capacity to create more value than the technology itself can deliver alone.

Failing to do so risks deploying a point solution, optimising a single process or aspect of the business, potentially slowing things down, creating more work, or deploying AI where the return is limited. MIT’s 2025 The GenAI Divide: State of AI in Business report, found 95% of firms produced no measurable P&L impact from their AI pilots, with failures attributed to poor workflow integration and organisational readiness. The lesson is not to abandon AI, but understand where it can be best deployed, and how.

Why must AI continuously evolve with the operating model?

In the mid-1950s AI was more symbolic, with humans hard-wiring rules and computers performing quickly against these rules. If a situation did not fit the logic, the AI broke. Now, AI determines the rules itself, predicting based on patterns it uncovers, and so applying this in organisations requires the right level of controls.

To avoid the ROI of AI diminishing over time, feedback loops, appropriate assurance, and evidence-based decision making must be built into the design of an organisation. AI is not a static piece of software and it can be impacted by market shifts, customer behaviour, the way data is used or controlled, regulations, and security vulnerabilities. It has the potential to erode trust with all stakeholders and ultimately destroy value.

Imagine a healthcare provider introducing an AI-powered appointment booking system to free staff time, and improve patient experience. The system required manual intervention, but the lack of formal feedback loops meant underlying issues went unaddressed, frustrating all.

Leaders must balance the choices they are making when it comes to AI and the associated trade-offs. Avoid investing in hype and allocate resources to where real value can be created.

How ready is your organisation for AI?

In one anonymised public sector pilot AI was used to automate part of a procurement process, including drafting documents ready for human verification. The pilot proved valuable because it exposed that the available infrastructure could not support the processing speed required to implement at scale. The organisation was not ready.

It’s an example where the ‘silver bullet’ solution generates excitement, and highlights that the readiness of the organisation could derail efforts.

There are 7 indicators that should be considered to maximise AI ROI.

IndicatorKey questionWhat you should do to maximise AI ROI
Strategic clarityAre we all clear on where we are heading?Determine what you want to achieve and be specific on how AI will deliver outcomes.
Leadership commitmentAre we running in the same race? Agree the parameters in which AI will (or will not) be used, and role model responsible use.
Organisational appetite for changeDo we want to do this?Demonstrate how working with AI, will improve outcomes and benefit staff, customers, and the organisation. Promises of transformative outputs are not enough.
Cultural readinessAre we able to change the way we behave?Create the environment in which use of AI can be challenged, questioned, and concerns raised. Resistance is not negative if it highlights improvement opportunities.
Change history and learningAre we holding to the past?Be explicit about past experience and learning of technology rollouts and change programmes, with a plan to mitigate any risk, improving confidence for users.
Capacity and capability to deliverCan we cope?

 

Invest in understanding and filling in the gaps relating to process, data sources, infrastructure, and the skills required to use the AI solution.
Operational stabilityWhat will distract us?An organisation in crisis or reactive mode risks creating more dysfunction with the introduction of AI. If in doubt, pilot.

These indicators will help you plan the best approach and respond appropriately, setting the organisation up for success when it comes to augmenting AI with the way it operates.

Conclusion

The organisations that maximise the ROI of AI will be those that understand their operating model well enough to identify where AI creates value and continue to redesign as work changes.

The ROI of AI is not achieved once implementation is complete. It is sustained by organisations that understand the whole system, embrace feedback, and adapt as requirements change. After all, no operating model is ever perfect or complete. 

About the Author

Matt KnibbsMatt Knibbs is author of The Grand Complication: A Guide to Organisation Design and Co-Founder of Kozo Design Solutions. He is a Certified Organisation Design Professional, Chartered Fellow of the CIPD and faculty trainer for the European Organisation Design Forum.

LEAVE A REPLY

Please enter your comment!
Please enter your name here