Why Most Companies Are Struggling With AI — And What the Numbers Actually Say

I want to talk about something that I think is really important. A months ago I was in a meeting and someone from leadership said that we should just add Artificial Intelligence to our project and that would fix all our problems. I did not say anything I just nodded along like everyone.

That moment really stuck with me. Because I had been reading about how Artificial Intelligence works inside companies and it is not that simple. So I decided to do some research and try to make sense of it all. This is what I found out.

Everyone talks about adding AI to fix problems — but the reality inside companies is far more complex. (Photo: Unsplash)

Artificial Intelligence is a big deal but the data situation is really messy

Every article I read about Artificial Intelligence talks about data. I used to think it was obvious that you need good data. Then I saw a statistic from IBM that said 42% of companies think that dealing with complicated data is their biggest barrier to using Artificial Intelligence. Not the algorithm, not the cost, the data.

Most companies have data spread across different systems that were never meant to work together. For example the finance team uses one system, the sales team uses another, the customer support team uses a third. Getting all of that data into a shape that Artificial Intelligence can actually use is a lot of work. It can take months to get it all sorted out before you can even start building anything.

Gartner says that bad data costs companies around $12.9 million per year. That sounds like a lot. When you think about it it makes sense. If you are making decisions based on wrong numbers, or if you have duplicate customer records, or if your reports do not match, it can add up fast.

Sources: IBM Global AI Adoption Index 2023 | Gartner, “How to Stop Data Quality Undermining Your Business” (2022)

Finding people who can actually do this work is really hard

When I was finishing my degree everyone was talking about Artificial Intelligence jobs like they were everywhere. In some ways they are. But the people who can actually fill those roles well are really hard to find.

Deloitte asked companies about this. 68% said that the gap in internal Artificial Intelligence skills was a major problem for them. Not a minor problem, a major one. The World Economic Forum says that Artificial Intelligence and machine learning roles are among the fastest growing globally. Demand is growing faster than people can be trained.

If you try to hire someone from outside the company you are competing with Google, Microsoft and every well-funded startup in the world. They are all chasing the same small pool of talent. Glassdoor says that senior Artificial Intelligence engineers can earn over $150,000 per year. A lot of businesses, especially mid-sized ones, just can’t compete with that.

Sources: Deloitte State of AI in the Enterprise 2023 | World Economic Forum Future of Jobs Report 2023 | Glassdoor 2023

The money question is one that nobody has a clean answer to

There is a big push inside companies to adopt Artificial Intelligence and budgets are being approved and projects are kicking off. But when you ask what the company actually got back from it, the room goes quiet.

KPMG surveyed executives and more than half of them said they struggle to show a clear return on investment from their Artificial Intelligence spending. MIT Sloan Management Review says that even a small Artificial Intelligence project can cost between $300,000 and over a million dollars once you account for infrastructure, data preparation and the team involved. That is for a small project.

PwC says that only 4% of companies have deployed Artificial Intelligence at scale with measurable financial results. 4%. The other 96% are either still experimenting, stuck in pilot mode or have quietly shelved the project.

Sources: KPMG Technology Survey 2023 | MIT Sloan Management Review | PwC AI Predictions 2023

When the room goes quiet after asking about AI returns — that silence is the data. (Photo: Unsplash)

Governments are catching up. Companies are not ready.

I did not follow Artificial Intelligence regulation closely until recently. It felt like something always ‘coming soon’ without actually arriving. But it is becoming a big deal now. The EU Artificial Intelligence Act is now in force. And it is not light reading.

Stanford’s Artificial Intelligence Index Report says that there has been a 26x increase in Artificial Intelligence-related legislation globally since 2016. That is not a typo. Twenty-six times more legislative activity in under a decade. Under the EU’s framework, using Artificial Intelligence for hiring, credit scoring or medical decisions gets classified as “high risk” which means you need audits, documentation, human oversight and more. For companies that were just plugging in tools without thinking much about it, this is a wake-up call.

Accenture found that 77% of business leaders are worried about Artificial Intelligence bias, which is the risk that an Artificial Intelligence system discriminates against certain groups of people without anyone noticing. What is more troubling is that only 35% of those leaders have a formal process for detecting or addressing it. Most are worried. But not doing much about it yet.

Sources: Stanford AI Index Report 2024 | EU AI Act 2024 | Accenture Technology Vision 2023

The technology is the easy part. People are the hard part.

I used to think that was a cliché. Then I started paying attention to how Artificial Intelligence projects actually fail in practice. It turns out it is just true.

McKinsey’s research on digital transformation says that around 70% of major change programs do not hit their goals. When you dig into why, it is rarely because the software did not work. It is because people were not brought along. Maybe leadership assumed everyone would just adapt. Maybe no one explained what was actually changing and why. Maybe employees felt like Artificial Intelligence was coming for their jobs and quietly pushed back.

Prosci looked at this specifically and found that projects with strong, deliberate change management built in from the start are six times more likely to succeed than ones where it is treated as an afterthought. Six times. That is not a small edge. That is the difference between a project that sticks and one that quietly gets abandoned after 18 months.

Sources: McKinsey & Company, “Unlocking Success in Digital Transformations” | Prosci Change Management Report 2023

Getting people on board matters more than getting the technology right. (Photo: Unsplash)

Most companies are trying to run Artificial Intelligence on ancient infrastructure

When you see an Artificial Intelligence demo it is running on clean data, modern cloud infrastructure and well-documented APIs. What most enterprises actually have looks nothing like that.

Accenture estimates that between 60 and 70 cents of every dollar in an enterprise IT budget goes toward just keeping existing systems alive. Patching them. Updating them. Making sure they do not fall over. That does not leave much room for plugging in something new and experimental. IBM’s research found that 40% of organisations listed integration with existing tools as a top technical challenge.

Sources: Accenture Technology Research 2023 | IBM Global AI Adoption Index 2023

So where does that leave us?

I am not trying to say that Artificial Intelligence is overhyped or that companies should not bother. I think it will be transformative for a lot of industries. But I do think there is a gap between how Artificial Intelligence gets talked about publicly and what actually happens when organisations try to implement it.

The companies that are getting results — that 4% — did not move fast and break things. They assessed where they actually stood, fixed their data foundations, trained their people, set realistic expectations and treated it like a long-term investment rather than a quarterly initiative. It may be boring. But it seems to be the thing that works.

ABICB helps organisations work through exactly these kinds of questions — from figuring out where you actually stand with Artificial Intelligence readiness to building a plan that makes sense for your specific situation. If any of this resonated, it might be worth a conversation.