Why Utility and Usability Drive AI Adoption

There's a certain narrative on the topic of AI that seems to be popping up a lot on LinkedIn recently. It goes something like this: If you don't adopt AI, your career is dead. You'll be overtaken by people who might not be as smart as you but are more proficient at using AI tools.

This kind of fearmongering is effective at driving anxiety (and clicks!), but a whole lot less effective at driving real behavioural change in the workplace. In fact, Nitro's latest Enterprise AI report of 1,000 professionals highlights a gap that shows up in most large organisations: 68% of executives feel pressure to deliver better AI tools, but more than half of employees—as many as 57%—feel little or no pressure to use them.

So why the gap?

Utility as a driver of AI adoption

Low AI adoption is not just a training problem or a lack of initiative. Both groups—execs and employees—have differing motivations. Executives are looking for cost savings, while employees are looking for utility. Can this AI tool really help me with my job? And if it can't, why would I change my process to use it?

To me, the gap suggests that many of the tools that execs expect will drive productivity and reduce costs are not actually proving useful in employees' day-to-day workflows.

This tracks with what I'm seeing in the market right now. Sadly, there are loads of software vendors out there either jumping on the AI bandwagon by simply slapping a generic AI assistant onto their existing product and hiking up the price, or trading on novelty rather than solving real customer problems. For example: Listening to your PDF as a podcast may be a neat use of AI, but it's not going to save you hours in your day or take lots of repetitive manual tasks off your plate. Some of these software providers have got too caught up in what AI can do and have forgotten about what their customers need to do.

The truth is that utility and usability drive adoption faster than top-down initiatives or a fear of getting left behind. High adoption will show up where value is immediate, the task is repetitive, and "smart" AI-powered tools fit naturally into a workflow without massive operational change.

Document handling offers valuable automation opportunity

Document handling workflows, which are often friction-heavy, check all the boxes. Employees spend hours every week on repetitive tasks such as extracting tables, turning documents into fillable forms, scanning for key details and protecting sensitive information before sharing documents—all of which can be automated in seconds. At Nitro, we've seen that smart tools can drive significant efficiencies when built well. Intuitive to use, reliable, and accurate are sine qua non. Our experience is backed up by our Enterprise AI report, which shows document automation has a meaningful impact: 89% of employees say AI saves them over 9 hours per week on these types of document tasks.

Product principles to drive AI adoption

To assess whether an AI feature will be shunned or embraced by users, I tend to return to four basic tenets:

1. Solve a specific existing user problem. Product development must start with a deep understanding of the user's workflow and pain points. AI should be a way to solve them, rather than a justification for building the feature.
2. Build trust into the product from the beginning. Deeply understand your customers' regulatory landscape. If security and compliance are afterthoughts, your product will never make it through enterprise governance and approval processes.
3. Make it intuitive, with little to no learning curve. Look for places you can remove friction or extra steps within an existing workflow. If users need to radically change their process to get value, you’ve created another barrier to adoption.
4. Ensure it works reliably and to a high standard. This may seem obvious, but one of the biggest hurdles AI has to overcome is the capability-reliability gap. It's not enough for a smart tool to produce good results sometimes; it needs to behave predictably and reliably every time, and the quality bar needs to be high from day one. A user will try it once and instantly judge whether it will be useful. You rarely get a second chance to impress.

These principles also help SaaS providers building vertical AI (tailored to specific industries or workflows) create more of a moat against the AI behemoths who offer foundation models. Claude and ChatGPT are by their very nature generic and flexible, but their strength is in breadth, rather than depth. By truly understanding customer use cases, regulatory environments, and pain points, vertical AI providers can provide what fintech investor Niha Bobba calls "last mile defensibility" by solving real complexity and adding a layer of trust.

Questions to ask when choosing the right smart tool for your organisation

If you're an executive feeling the pressure to select and roll out AI tools that genuinely drive productivity and cost savings for your organisation, you can pressure test likely adoption early with a few practical questions.

• Does this remove or speed up one or more steps of an existing repetitive workflow?
• Is the tool intuitive enough that a user get value from it without a tutorial?
• Can the output be reviewed easily?
• Has security and compliance demonstrably been considered in the design of the product? Detail on this topic should be readily available from the vendor and published on their website.

As an exec, enabling and streamlining approvals for tools that meet the criteria above helps you realise productivity gains sooner and reduces risk for your organisation. Nitro's Enterprise AI survey indicates that where these needs are not met through official governed channels, unapproved AI use is common, with a third of employees saying they have processed confidential company information through AI tools.

You don’t need to be scared into using AI

The potential to achieve better outcomes is a more effective “carrot” to drive the use of AI than the “stick” of top-down mandates or a fear of getting left behind. The simplest test is to try using a tool as part of your regular workflow and see what happens: Is it just another snazzy AI feature, or is it a smart solution to an everyday problem? If it speeds up a task and makes your job easier, it’s likely to earn its place as your default option organically. My view is that the rate of AI adoption, like any other technology, will ultimately be determined by utility and usability.

Originally published by: Cassie Harman, Chief Product Officer, Nitro