Why relevant AI decides the future of work

A year ago, a financial forecast or a report into a customer bug find meant an email to the relevant team, a disruption to their work and a delay for me. Now, I ask an AI assistant, point it at the right numbers, and have the answer before the kettle has boiled. It’s a fairly small change in the era of AI, but small changes, repeated across every desk in a company, are how the fabric of work is being reshaped.

Summary: A foundation of relevant AI data is the trusted layer of company information an AI draws from. It, not raw model intelligence, decides whether AI gives people real leverage or fast, confident errors. Most AI projects fail because that layer is missing. Build it well, and every answer compounds.

How is AI redrawing the lines between roles?

AI is letting people do the work of three or four roles at once. The walls between jobs are getting thin; in some places they are already gone. This is good. It puts capability back in the hands of the person closest to the problem, and makes that person faster, more knowledge-rich and ultra-efficient.

Watch any team for a week and you’ll see it. The CEO who now runs their own financial analysis in real-time, the QA engineer who fixes the bug that once upon a time she could only flag, the Customer Success manager using AI with a client on-site to explore complex processes instead of raising at ticket. I’ve started calling this person a blended specialist. It’s not a job title; it’s description of what a lot of us have become.

This isn’t just a hunch. McKinsey’s 2025 State of AI survey found that 88% of organizations now use AI in at least one business function, and 72% use generative AI, up from 33% a year before. The functions where AI agents are most impactful are IT and knowledge management, the exact places where work used to be able to fall through unintended cracks or be handed off to numerous colleagues. Moderna folded its HR and IT leadership into a single role, and readily admitted what for many is taboo: AI is no longer a tool, it is a colleague. We go deeper on how AI is changing the workplace elsewhere.

Why relevant AI that can act immediately matters more than a clever model

“A model cannot tell good information from old, stale information. It knows one thing: what it can reach.”

None of this works simply because the machine is clever. It works, when it works, because the machine can reach knowledge that’s worth trusting. An AI knowledge base is what makes that possible: one reliable place the intelligence draws from.

Go back to the financial forecast that I wanted. The answer was only ever going to be as good as the numbers propping it up. Scatter those numbers across a dozen systems, let half go stale, let a load of it contradict each other, and your AI tool of choice will still answer. It will answer instantly. It will answer with total confidence. And it will be wrong. A model – any model – cannot tell good information from old, stale information. It knows one thing: what it can reach.

That changes the question I care about as Interact’s Chief Executive. The future of work is not a story about how smart the AI gets. It is a story about whether the company underneath it is worth asking. Whether the data feeding your AI is trustworthy. Whether everything you’re building towards is terrifyingly good, or terrifyingly bad.

Get that right, and you hand every person extraordinary leverage. Get it wrong, and you hand every person confident nonsense, faster than they have ever received it before. The same AI with the opposite outcome. It’s why strong enterprise knowledge management is fast becoming the deciding factor in many organizations’ AI roll-out.

What does the data say about why AI projects fail?

The data says that the technology is rarely the bottleneck. The knowledge underneath it is. MIT’s Project NANDA studied hundreds of enterprise AI deployments through 2025 and found that 95% of organizations got zero measurable return from generative AI. Zero. Gartner expects companies to abandon 60 percent of AI projects through 2026 for a single reason: the data underneath them was not ready. In a Gartner survey of infrastructure and operations leaders late last year, nearly four in ten named poor or missing data as a direct cause of failure.

McKinsey tells the same story from the other end of the table: roughly eight in ten companies use generative AI. Roughly eight in ten companies report no difference to the bottom line.

“Roughly eight in ten companies use generative AI. Roughly eight in ten companies report no difference to the bottom line.”

Those numbers say one thing clearly. It’s the same operating model, but with two entirely different layres, delivering opposite results.

Dimension	A base of relevant and trusted AI	A fragmented and noisy AI tool
Source of truth	One version	Mulitple, often contradictory
Speed of answer	Fast	Fast
Accuracy of answer	High and traceable	Confident but often wrong
When a person leaves	Knowledge stays	Knowledge often leaves with them
Business outcome	Compounding leverage	Wasted spend, abandoned pilots

The speed is identical, but everything else of value is anything but.

What does trustworthy AI actually look like?

A trustworthy AI knowledge base is not “just one more thing to bolt on to our platform.” It is one version of the truth instead of twelve. It’s the right answer made easy to find and the wrong one retired. It’s knowledge that you can build on, without worrying who will take it with them when they move on.

“If the foundational of your structure aren’t sound, you’re building an expensive, slick, powerful tool that may deliver disastrous results.”

In practice, the layer beneath the results has a few unglamorous qualities:

• One source of truth, not a dozen systems delivering a confused and non-linear set of answers
• The current answer surfaced and the stale one retired
• Institutional knowledge captured, so it survives the people who hold it

This is the Knowledge pillar of employee experience doing its crucial job, and it is the only thing that decides whether the proecss sitting on top is actually worth anything. It is exactly the work we do at Interact: not just the shiny toy, but the layer beneath it, the enterprise employee knowledge base where an organization’s knowledge comes together so the answer is the right one.

It’s very easy to be carried away on a wave of AI excitement, but if the foundational components of your structure aren’t sound, you’re building an expensive, slick, powerful tool that may deliver disastrous results.

How can leaders build a trusted and relevant AI core that compounds?

Leaders build it by treating knowledge as the foundation. This is the part that hands the advantage back to you. You will not win the next few years by buying the biggest model. At this point, everyone has access to power and potential unlike anything we’ve seen before. You win by running a company whose knowledge is good enough to be entrusted to AI.

A practical order of operations:

1. Consolidate before you automate. Collapse multiple scattered sources into one. Relevant AI knowledge built on embedded AI for employee experience is only as good as the sources feeding it.
2. Govern what the AI can reach. Decide what is current, retire what’s stale, and make ownership clear. Good enterprise knowledge management is governance, not guesswork.
3. Make the right answer easy to find. Pair the knowledge base with strong enterprise search so people and the AI beside them land on the same trusted answer.
4. Keep it true over time. Every answer you keep accurate makes the next one faster to settle. That is the compounding part, and it is yours to build.

The future of work starts with a core of relevant AI tools

This blended workforce is coming, whether we are ready for it or not. The only question is the one you can actually answer: When employees enter their blended era, will the knowledge inside your company be ready to meet them?

Long before arguing about which model to buy, you need to get the knowledge right. At that point, the future of work stops being something that happens to your organization and becomes something you do better than anyone else. If you’re ready to build a platform that delivers rtusted, reliable and relevant AI your people and you can trust, our people are waiting to start the conversation.