The Dunning-Kruger Effect with AI in communications is not what you think

Limited AI skill may be making communications professionals blind to their own gaps

Feeling more confident about how you use AI?

The rising confidence many communications professionals feel about artificial intelligence is not hard to explain. The tools have improved quickly. They are easier to use, more widely available, and increasingly embedded in the apps people already rely on, including search, email, documents, meeting notes, and the daily systems of work.

But ease of use is not the same as skill. With generative AI, that distinction is becoming more important and more easily missed.

For the past two years, Sequencr AI has worked with communications teams to empower them to understand, adopt, and scale the use of the technology. At the start of every engagement, we ask participants how they use AI, how proficient they believe they are, how confident they feel about their skills, and where they hope to apply the technology next.

The answers point to a troubling pattern. Communications professionals are becoming more confident with AI, but they are not becoming more proficient. In fact, proficiency has stagnated.

When confidence outpaces capability

Across dozens of Sequencr AI surveys, with more than 1,000 substantive responses, self-reported proficiency has not changed significantly over the last two years. The largest group continues to sit in the middle. In surveys where we asked a comparable proficiency question, 56 per cent of respondents identified themselves as Explorers, one level above beginners. Another 20 per cent identified as Adopters, 12 per cent identified as Skilled Users, and just two per cent qualified as Power Users. 

Confidence tells a different story. In recent surveys, it has risen from an average of 4.2 a year ago, to the mid-6s on a 10-point scale.

In some teams, we are seeing a clear divide. Half of the people with similar proficiency scores reported low confidence in their AI skills. The other half rated their confidence at 7 or 8 out of 10, even with the same reported proficiency.

People are more confident of their AI skills even though their proficiency has not changed. 

This dynamic resembles the Dunning-Kruger effect, the cognitive bias in which people with limited knowledge or ability in a specific area overestimate their competence.

For communications teams and agencies, that confidence gap has consequences. It affects what teams try, what they ignore, and what risks or opportunities they fail to see, including:

1. Teams mistake basic use for real capability

Most communications professionals already use AI to draft, edit, summarise, brainstorm, and rewrite. Those are useful applications. They are also the easiest ones.

The larger opportunity is not to make the old workflow a little faster, it is to rethink the workflow itself. AI can help teams test messages across audiences, map stakeholder concerns, identify reputational risks, synthesise competing signals, draft response scenarios, and adapt content across channels with far greater speed and consistency.

A team that believes it is already proficient may never ask the more important question: What could this work become if AI were designed into the system, not simply added to the task?

That is missed opportunity.

2. Teams let AI do the thinking they still need to own

Overconfidence changes how people use AI. When teams believe they are more proficient than they are, they are more likely to accept the tool’s framing, structure, and recommendations without challenge.

That is where confidence becomes a liability. Teams may feel more capable because they are producing more, while becoming less practiced at the thinking that creates real communications value.

3. Teams skip the context layer

Overconfident users often assume the problem is the prompt. If the output is generic, they try a sharper instruction. If it lacks nuance, they ask for another version. If it misses the audience, they rewrite the ask.

But communications work depends on context that a generic prompt cannot supply on its own: the organisation’s history, stakeholder relationships, source environment, issue dynamics, risk tolerance, brand voice, and strategic priorities.

Without that context, AI produces plausible work, not useful work.

That is why the next level of proficiency is not just better prompting. It is building the systems, knowledge bases, and workflows that give AI the right information to work from.

4. Teams fall behind the competitive curve

The biggest risk of misplaced confidence is complacency. If teams believe they are already good at AI, they are less likely to keep pace as the technology changes.

AI is no longer just a prompt box. Agents, automated workflows, and connected tools are beginning to handle more complex tasks: monitoring signals, retrieving context, drafting outputs, and coordinating multi-step tasks.

Teams that overestimate their proficiency may keep using AI in basic ways while assuming they are keeping up.

In a fast-moving technology cycle, that is how confidence turns into competitive disadvantage.

The real confidence issue with AI

That is why the Dunning-Kruger effect with AI in communications is not quite what people assume.

The obvious concern is that AI makes people overconfident about other subjects. Someone asks a tool to explain a legal issue, a financial trend, or a policy debate, then mistakes a fluent summary for real expertise.

Our data points to a different problem.

Communications professionals are not necessarily becoming overconfident because AI makes them feel like experts in everything else. They are becoming overconfident because AI makes them feel more capable at using AI itself.

Our proficiency data suggests there is more to learn and more upside to take advantage of.

Matt Collette is CEO of Sequencr AI, a technology consultancy dedicated to unlocking the full potential of AI for marketing and communications teams. Before founding Sequencr, he was Head of Digital for Edelman Canada and later Global Head of Digital Growth, where he led efforts to embed generative AI across Edelman's global operations.

Matt created the firm's AI task force, launched its first campaign powered by generative AI, and developed tools, prototypes, and training initiatives for clients and internal teams.

Read more from our columnists in The Earned View