How do you turn AI into business value?

There is a general consensus amongst the major consulting firms that whilst investing in AI isn’t worthless, they recognise that over 80% of AI project initiatives fail to scale beyond the pilot.

So what’s happening?

Why isn't AI delivering value?

An AI report from advisory firm Bain and Co. highlights the misalignment between what the leaders expect and what AI is delivering. Moreover this mismatch will eventually grind the AI investment to a halt.

The misalignment is in fact a conflation of productivity and business value. It also touches on the systemic tension between the IT function and the c-suite. The latter often prefers to keep the former at arm’s length and so the IT function becomes the victim of business value initiatives rather than its co-creator.

Thus the IT function’s focus veers towards operational efficiency / productivity:

• Hours saved

• Faster processes

• Greater automation

• Increased uptime / business continuity.

All important. But business leaders care more about matters such as:

• Revenue

• EBIT

• Margin

• Customer retention

• Market share.

AI is delivering on the former, but it is not so clear about the latter. Hence the overall sense of disappointment. More broadly, organisations are designing AI to optimise yesterday's stable world, while businesses need AI that helps them navigate today's disruption, uncertainty and complexity.

How do you measure AI ROI?

But perhaps all parties are missing the key point. Measuring AI solely in respect of its financial return on investment is to not fully understand the value that AI can deliver.

Beyond efficiency, we need to consider:

• Capability ROI – Can we now do things that were not possible before the AI investment?

• Strategic ROI – What happens if we do not make the investment?

• Cognitive ROI – How will this improve the organisation’s ability to sense, decide and act, and thus adapt to a rapidly evolving reality?

Efficiency ROI is a relic from the industrial era. A time when leaders, investors and analysts considered profitability to be paramount.

Cognitive ROI is critical. In an increasingly disruptive world, an organisation’s ability to operate in real-time is critical to its survival. And perhaps the big reveal is that your organisation is already heavily invested in this respect. ‘Employer brand’ is a strong measure of how well you have nurtured this asset. But now AI can be applied to bolster this asset class.

In the medium term at least, it is worth focusing on how AI can augment your human investment and thus boost not just productivity, but also the organisations's innovation capacity.

Why do AI projects fail?

Again inflated expectations have a role to play. The tech marketers have created a vision of the enterprise as a silent disco - an organisation comprising autonomous AI agents sliently turning inputs into profit.

Governance layer

We are some way off what would essentially constitute a post-human economy. Only someone who cares little for governance would allow agentic AI to roam freely within, and beyond, the organisation. As an aside, better to think of agentic AI as a new employee with good credentials. Even so, it is best not to assume they will do a good job. So either mark its homework (human in the loop) or monitor from a distance (human on the loop).

Workflow layer

But even if the governance layer is in place, there is still the issue of using AI to faithfully replicate a flawed business process.  There is no point allowing AI out of the lab if your organisation is blighted with workflow debt.

Data layer

But let’s assume you now have an elegant set of business process flows. It would be unwise to unleash workflow AI if your enterprise data was more akin to a cesspit than a pristine data lake. What is not yet widely understood is that AI is excellent at data cleansing. The IT function would be wise to explore this approach rather than trying to perform major systems integration surgery.

The bottom line: Deploying AI before the prerequisites are in place will not end well.

Concepts aren't reality

IBM, in an article entitled ‘Why AI projects fail, highlighted that a major cause of failure is treating AI proof of concept activity as isolated science experiments conducted in a sanitised ‘lab’. These fail in large part because the primary stakeholders were not involved from the outset. Stakeholders include users and the CEO. Appointing a Chief AI Officer to oversee this simply delays the realisation that the pilots are not going to scale.

AI done well

Financial advisory firm, The Motley Fool compiled a list of 10 organisations that are using AI to great effect.  Examples include:

• JP Morgan Chase – Fraud detection, research and credit risk management.

• Amazon – Warehouse management and demand forecasting.

• Netflix – Hyper-personalisation of its service.

• Johnson and Johnson – Drug discovery and improved surgical precision.

These are not headline grabbing applications. But they are creating real business value. The media, AI keynoters and AI conference organisers take note and tone it down a little.

What organisational capabilities determine AI success?

Trust

This is key to successful AI adoption. If your people are fearful, they will act as rogue organisational antibodies primed to thwart AI adoption.

Assets

Leaders need to think less about efficiency and profit (industrial era notions) and more about business value (ie growing assets). Today, very little attention is given to assets beyond cash / cash equivalents and physical assets such as property and machinery (which are not even assets). I would strongly encourage organisations to orient around three emerging assets:

• Cognition – Nurture your people and augment them with AI. The investment distribution between AI and people needs a revisit. Its time to sit down with the CHRO and the CIO.

• Stakeholder – The relationship with all stakeholders from investors to the planet need to be managed with care. Well managed relationships create value.

• Governance – An increasingly disruptive world, fuelled with AI, needs a new type of governance. If traditional governance is a baby crawling, today’s government is akin to a high-performance stuntman.

What does an intelligent organisation do differently?

Intelligent organisations are focused on staying in the game, ie survival. No matter how mature your organisation, each day is day one. And each day should include a healthy dose of both fear and excitement. Delivering business value in an increasingly uncertain world requires youthful agility.

Intelligent organisations don’t just follow the crowd, they do the foundational work first, including:

• The organisation design work

• The pre-deployment tenderisation work.

Avoiding these will create more problems than AI solves. Turning a sausage machine into an adaptive living system requires more than building/buying an LLM and strapping it on to the frame.

Imagine your organisation is a washing machine. If you are not careful, adding AI to your organisation is the equivalent of using sand for detergent and throwing loose coins and keys into the drum.

If you ignore these foundational points, at your next shareholder meeting you will find it very difficult to put a positive spin on your AI investment.