Pressure to adopt AI isn’t coming from one source. It’s coming from boards, competitors, customers who have already experienced AI-powered service elsewhere, and employees who increasingly expect AI to make their work easier.
But pressure and clarity are two different things. For many mid-market organisations, the reality is that they’re navigating a landscape that’s moving faster than their ability to evaluate it thoughtfully.
The risk of moving too quickly is real. So is the cost of waiting too long.
Mark Sher, SVP of Product Marketing at Intermedia, sees this tension every day.
His view is that AI adoption in business communications isn’t happening all at once. Organisations are moving at different speeds and for different reasons.
“Some say, ‘I’ve got to have it right now. I don’t care,'” he says.
“Others are more pragmatic – they want to understand the ROI, they want to understand what they’re going to get in return. And there are others still crossing their arms and saying, ‘I’m not ready.'”
The pragmatists, Sher believes, are asking the right questions. The organisations seeing the strongest early results are typically those that begin with practical, well-defined use cases before expanding into more advanced automation.
Not All AI Is the Same – and the Distinction Matters
One of the most useful frameworks Sher offers is a simple distinction between two types of AI. Understanding the difference, he says, helps organisations avoid many of the pitfalls that accompany early AI adoption.
The first category is what he describes as out-of-the-box AI – technology that delivers value almost immediately with little configuration or training.
Meeting transcription is a good example. The meeting ends, a summary is generated, action items are captured and a record is immediately available. These capabilities require minimal setup, produce immediate productivity gains and are among the lowest-risk, highest-return AI capabilities organisations can adopt.
Call summaries follow the same pattern. In a contact centre, an agent finishes an interaction and AI automatically summarises the conversation and updates the CRM.
“I don’t think there’s a ton of risk with that,” says Sher. “And there’s a lot of upside – around productivity, efficiency, a better employee experience and a better experience for the customer.”
The second category is knowledge-dependent AI – and this is where both the opportunity and the complexity increase.
An AI voice agent that can answer calls, hold natural conversations, schedule appointments and intelligently route enquiries is a compelling capability.
Its effectiveness, however, depends entirely on the quality of the information it’s given.
“To make that work right, you have to invest in infusing it with the right kind of business data,” says Sher.
“A lot of it comes down to your ability to compile your business data in a useful way for the AI.”
AI Is Only as Good as the Information Behind It
That point is worth dwelling on because it reframes the AI conversation in a way that’s particularly useful for mid-market businesses.
More often than not, the technology isn’t what holds organisations back. Their data is.
Preparing accurate business knowledge, policies, customer information and workflows often becomes the biggest part of a successful AI initiative.
“There’s a saying with computer programs – garbage in, garbage out,” Sher says.
“You want to make sure that the information about your business is up to date, accurate, compiled well, and then provided to the AI so it can do a really good job. If you don’t give it good data, it can’t do a really good job for you. And that’s when some of these initiatives fail – usually it’s the data, not the AI.”
For mid-market businesses without dedicated data teams, that’s an important consideration before committing to any AI implementation.
The question isn’t simply whether the technology works.
It’s whether the business is positioned to make it work.
Why Platform Matters
Another challenge organisations face is deciding where AI should live.
Many instinctively bolt AI onto existing systems rather than adopting a platform where AI is built in from the start.
The appeal is understandable. It feels like a lower-risk way to introduce new capabilities without replacing existing infrastructure.
In practice, however, it often introduces additional administrative overhead instead of reducing it.
“Bolting anything on usually means it’s not integrated,” says Sher.
“If you have multiple different bolted-on solutions, each one is going to require data. Now you’ve got various data sources that you have to continue to update and manage in all of these different places.”
His recommendation is straightforward: start by looking at the vendors you already trust and determine whether AI has been integrated natively into their platform.
A fragmented technology environment often leads to fragmented AI, where different assistants produce different answers because they’re drawing from different sources of information.
Digital Teammates – What That Looks Like in Practice
Intermedia describes its approach as AI teammates – assistants that support employees in their day-to-day work and agents that can perform specific tasks autonomously.
The practical impact is particularly evident in the contact centre.
AI that listens to conversations in real time and surfaces relevant answers as questions arise can have a measurable effect on agent performance.
For business leaders, that shifts AI from an innovation project to an operational improvement with measurable business outcomes.
Read more: Intermedia CEO Mike Gold on its 26North Acquisition, What’s Next, and More
