The $1.6 Million Lesson: Why AI Risks Becoming the Next Shadow IT

AI / AN / AO (Artificial Intelligence, Autonomous Networks and Autonomous Operations) transformations create different challenges for different stakeholders, depending on where you sit in the org chart of a network operator. Business units, technology teams, operations, architecture, finance and governance all see the autonomy problem through a different lens.

This week, I’ll be sharing an article each day with a different, but ultimately converging, perspective on the AI / AN / AO transformation challenge.

Then we’ll bring them together into a consolidated take-home pack you can use to help shape your own AI / AN / AO transformation planning.

Let’s start with a problem we faced with a client. It was to solve a business problem: how could we make the contact centre more useful for customers and for the telco itself?

Pre-AI chatbots were one of dozens of different initiatives we investigated as a means to solve the business problem. In investigating chatbots we came across a really interesting conundrum that’s arguably even more applicable in the AI world today.

Like many of our projects, we started by talking to people across multiple business units. In doing so, we discovered something nobody else seemed to have been aware of.

It turns out that seven or eight different business units within our client’s organisation had already investigated chatbots.

Each had spent roughly $200,000 on consulting and investigation during the previous six months. As it turned out, none of the units had enough funding to move from investigation to implementation. Each had done the investigation, but none had built the foundational capability. And it turns out that none were aware that the others had been exploring the same opportunity.

.

The duplication wasn’t the biggest problem

Naturally, you can do the maths. If 7-8 teams spent around $200,000 each, that means the organisation had spent $1.6 million investigating broadly similar ideas.

If all of the teams had’ve pooled their budgets, $1.6 million would have got them the foundational capability that each could’ve then leveraged…. But by the time we’d identified it, the $1.6 million was spent and there was no more budget to do any form of implementation.

The team we were representing, effectively team number 9, could have easily have started again from scratch, rebuilding the same understanding around data access, integration, security, governance, model choices and connections into OSS and BSS. Unfortunately for our project, chatbots didn’t really move the needle as much as about a dozen other initiatives, so we never built a chatbot for them either (nor did we spend another ~$200k investigating them).

But that is a pattern etched in my brain now. I see something really similar today.

Plenty of AI activity, but very little shared AI capability being built across teams within the same organisation.

.

This is how shadow AI emerges

In that particular case, each business unit understood its own customers, objectives and constraints. Each saw potential value in the chatbot initiative and commissioned work it could justify within its own boundaries.

Individually, those decisions made sense. But when we took an outsider’s more holistic perspective, they created an outcome nobody would have designed deliberately.

This has some similarities with the pros and cons of what’s known as Shadow IT. Shadow IT is the use of technology, software or services by business units with their own budgets or resources, that are purchased or acquired outside central IT governance. This often helps the business unit to move faster but potentially at the cost of visibility, consistency and enterprise coherence.

I’m not suggesting shadow IT is good or bad, just that it exists with strengths and weaknesses.

AI development risks becoming the next generation of shadow IT, with corresponding pros and cons.

The danger isn’t simply that several teams buy the same thing. It’s that they spend money learning the same lessons without creating anything that other teams within the organisation can reuse.

.

The funding boundary may be the problem

There’s another question hidden inside the $1.6 million chatbot example.

What if no individual business unit can justify funding the shared capability, but the organisation as a whole is already spending enough to build it?

Many AI foundations create value across organisational boundaries. A capability created for the contact centre today might support service assurance, network operations or sales tomorrow.

If every investment has to justify itself within one business unit, we risk underfunding the things everyone needs while repeatedly funding investigations into them. Or assigning resources to smaller projects separately, which means smaller objectives and outcomes can be achieved, when bigger collective goals / outcomes might be more valuable.

.

AI needs coherence, not central control

The answer isn’t to stop business units experimenting. Local teams will almost always have the best understanding of their own problems and opportunities. They certainly have the best context.

The challenge is making sure those investments can roll up and contribute to something bigger.

The biggest risk from shadow AI isn’t that five teams buy the same thing. It’s that five teams spend money learning the same lessons without building anything a sixth team can reuse and build upon further. Standing on the shoulders of giants!

In the next article in the series, we’ll look at why everyone wanting their own AI project is both an enormous opportunity and a potential obstacle to building a coherent AI capability.

Download our AI / AN / AO flyer for an overview of the challenges we’ll explore throughout the series. At the end, we’ll consolidate the lessons into a practical take-home pack you can use in your own transformation planning.