AI discernment in business communications: AI gives you options, expertise makes the decisions

AI discernment in business communications: AI gives you options, expertise makes the decisions

TLDR: AI discernment in business communications is the difference between using AI to raise a floor and letting AI flatten a ceiling. AI works on the law of averages, which means it can pull below-average work up and quietly pull professional work down to the middle. What separates authoritative brands from generic ones is not access to AI – it is the expertise to know which options to accept, which to reject, and when to stop.

Most business communications teams now have AI somewhere in their workflow. Drafting supplier emails. Rewriting proposals. Summarising meetings. Reworking newsletter copy. The tools are fast, cheap, and – if the demo is anything to go by – impressive. And yet a strange thing is happening across a lot of the output landing in inboxes and on client desks. It all sounds a bit the same.

Stanford’s 2025 AI Index reports that 78% of organisations used AI in 2024, up from 55% the year before. Adoption is not the question anymore. Discernment is. The teams getting real value from AI are not the ones using it the most – they are the ones who have decided what to use it for, and just as importantly, what not to use it for. That decision is a professional judgement, not a prompt.

This article is a thought leadership piece written by Louise Cunningham, co-founder of CommsKit, about the two failure modes we see most often – the law of averages problem and the infinity loop problem – and what changes when experienced people apply AI discernment in business communications instead of outsourcing the thinking to the tool. For a companion piece on where AI actually earns its keep inside a comms function, see AI in communications: strategy first, tools second.

What is AI discernment in business communications?

AI discernment in business communications is the professional judgement to decide when AI improves the work, when it degrades the work, and when to stop asking it for more. It sits somewhere between prompting skill and editorial expertise. It is what stops a competent brand voice being sanded down into something that could have come from anywhere.

The reason this is a live issue is simple. MIT’s Project NANDA State of AI in Business 2025 report found that despite $30–40 billion of enterprise investment in generative AI, 95% of organisations are getting zero measurable return. The tools are not the problem. The way they are being used is.

The law of averages problem

AI has democratised competent writing and design. For organisations starting from a low base – a first-time drafter, a stretched team, someone writing outside their comfort zone – that is good news. AI will pull the work up. The floor rises, and that is a real gain worth having.

But professional brands do not compete by being competent. They compete by being distinctive. Left to its own devices, AI drifts towards the language, ideas and designs it has already seen a thousand times: harmless while a team is starting out, corrosive the moment the baseline is already professional. That is the gap AI discernment in business communications exists to close.

For a senior comms lead, an experienced account director, or a specialist with fifteen years of sector knowledge, AI pulls the work toward the middle instead. Not badly. Not obviously. Just enough that the distinctive edges get rounded off, the specific becomes general, and the piece that would have been unmistakably yours starts to read like any other agency’s output. The ceiling drops.

AI can both raise the floor and lower the ceiling. How you practice AI discernment in business communications defines how if it boosts or limits you.

Take a message that has to go out to the whole company – a restructure, a change to shift patterns, a safety incident that needs explaining. AI can turn a rough first draft into something clear and well-structured in minutes. If you are not a confident writer, or you are staring at a blank page under time pressure, that is a real win.

But a message like that is not just information. It has to carry the reader’s trust – in the leader whose name is on it, in the tone the organisation uses when things get difficult, in the specific history of that workforce and what they have been through before. AI does not know any of that. Left unchecked, it writes towards the average of every similar announcement it has been trained on, and it will happily produce something calm, professional and reasonable that could have gone out from any organisation, about any restructure, to any workforce. The version that would have actually landed – because it sounded like the leader people know, and it spoke to the specific moment that workforce was in – gets rounded off into something generic. The ceiling drops.

That is why human expertise, brand knowledge and critical thinking matter more than ever. The role of professionals is no longer simply writing the message. It is deciding what AI’s version gets wrong about this audience, and fixing it before it goes out. Discernment is the filter that decides which suggestions raise the standard and which ones lower it.

The infinity loop problem

The second failure mode is less discussed and, commercially, more expensive. AI will iterate forever. There is always another version, another rework, another “sharper” pass. Without a stop rule, teams disappear into revision cycles that cost hours and improve nothing.

A short scenario we have seen play out more than once: a comms lead is drafting an all-staff email about a change to the shift pattern. First AI draft: clear, but a bit direct. Ask for warmer. Second draft: warmer, but now vague. Ask for more direct. Third draft: direct, but the tone has drifted from how leadership normally talks to the business. Ask for a rewrite in the organisation’s voice. Fourth draft: closer, but a paragraph about timing has crept in that was not needed. Fifth draft: trimmed, but now it reads like every other policy email people have had this year. By draft six, thirty minutes are gone, the email is worse than draft one, and it sounds like every other update the workforce will read that quarter.

The recent Theta Lake / Thomson Reuters research on AI-generated communications found that 88% of firms are already struggling with AI governance, and the top challenge – cited by 47% – is ensuring AI-generated content is accurate and compliant with regulatory standards. Note what is not on that list: “generating more of it”. The bottleneck is judgement, not volume.

The infinity loop matters commercially for two reasons. It spirals time and cost on work that should have taken minutes. And it dilutes the brand voice one polite, plausible rewrite at a time. If AI is drafting your responses to partners and suppliers, and no one is deciding when to stop, you are paying twice: once for the time, and once for the reputation.

What does AI discernment in business communications look like in practice?

Discernment is not mystical. It is a small set of decisions made consistently, by people who know the work well enough to make them.

6 tips for more effective AI discernment in business communications

Decide what AI is for, before you open it. Repetitive, structural, low-judgement tasks – meeting notes, first-draft scaffolding, tidying grammar. Not final client-facing copy in a sensitive relationship. Our guide on how to build an effective AI workflow for internal comms sets out a simple four-step way to draw that line.

Set a stop rule. Two rewrites, then the human takes over. If you cannot get it right in two passes, the problem is the brief, not the model.

Match the seniority of the reviewer to the sensitivity of the audience. A supplier email in a strained relationship needs a senior eye. A monthly newsletter does not.

Protect the distinctive parts of the voice. The phrases, cadences and specifics that make your brand yours are exactly what an averaging engine will smooth out. Flag them and keep them.

Treat AI output as options, not answers. The tool gives you three possible directions. Your expertise picks one, or discards all three.

Do not use AI for the parts of the work that are the reason the client hired you. If judgement is what they are paying for, it is the last thing to automate.

These are not rules that a tool can enforce. They rely on people who have done the work long enough to know when the machine is helping and when it is quietly making the work more average.

Where CommsKit sits on this

CommsKit is the evolution of Halo Media, an internal communications practice that has been delivering work in complex, regulated and workforce-heavy environments since 2008. That track record of judgement calls about when a message lands and when it does not is what shapes how we use AI on client work today.. We use AI as an internal support tool, governed by our Responsible AI Use Policy, and every output is reviewed and approved by a CommsKit team member before it leaves the building. AI expands what our team can do in a day. Experience decides what actually goes out. This is also the thinking behind our own AI work in the sector – see building an AI that actually understands comms work for how that shows up in practice.

That is the working principle behind the CommsKit approach to AI: options from the tool, decisions from the people who have done the work at scale.

The takeaway

AI is a good floor and a bad ceiling. Used with discernment, it saves time and lifts standards. Used without discernment, it costs time and lowers them. The organisations getting a real return on their AI spend are the ones who have decided, in advance, what AI is for and where the human decision sits. That is what AI discernment in business communications comes down to in practice.

If you want to see how that principle plays out inside a managed communications service, talk to our team to set up a discovery call.