AI: The Reputation Signal · Issue 5
Slow thinking, out loud
Elif Güvençer ·
Most inventions result from a personal need or frustration. The Two Clocks Framework™ — where I tackle what AI means structurally for Comms and how the function needs to reposition — is the result of frustration. A frustration with most of the conversations available to Comms being tactical. Monitoring dashboards. Content automation. Generative engine optimisation. Useful, but insufficient as a primary response to AI.
I have recently finished a book that also addresses what AI means structurally for Comms: AI for PR, edited by Stephen Waddington and Ben Verinder. Beyond being a rich resource with well-considered opinions and expert recommendations, its ripple effect — what happens as the field reads it — is part of what makes it valuable.
There needs to be more conversation, more thinking, more talking about this topic to define and address not just what AI means for Communications, but what Communications means for AI.
The discipline has a role to play in ensuring that AI, from conception to deployment, is developed ethically, responsibly, and in a socially sound way. It is sufficient to look at AI-related news to see how much this role is needed: massive layoffs, security concerns, fake content, commencement speeches being booed…
In the last newsletter, I named Communications as the organisation's slow thinker. Reading the book pushed me to ask where else that role applies — beyond the organisation and into the public conversation about AI itself. Four areas came to mind. They are not the ceiling.
1. The truth layer
I've written about eligibility as the AI-era reputational question: organisations can be visible to AI answer engines and still not make the cut. Companies that appear on an AI-mediated vendor shortlist are those that are eligible- a function of consistent, authoritative signal architecture.
Nate B. Jones gives that idea executional shape in communication context, framing it as the truth layer AI systems need underneath brand narrative. When AI systems assemble shortlists, they weigh verifiable, structured signals over brand narrative. "Best shoe in the world" might land with a human; it does not survive a process that seeks substantiation beneath the claim.
Comms' due diligence muscle — sourcing, fact-checking, and proof points — is ideal for contributing to that truth-layer.
The societal extension matters more. If AI systems scale information to everyone who asks, what gets fed in becomes a public concern, not just an organisational one. Communications, serving as the truth layer of organisations, has a role beyond corporate boundaries — helping shape what is accepted as input to systems that compress and distribute information at large. That is slow thinking applied to public discourse, not just organisational reputation.
2. Environmental literacy
It took years for environmental sustainability to progress, from regulatory concern to public consciousness. Good communication did much of that translation work. Carbon footprint calculators. Hotel cards translating minutes in the shower into gallons of water. Recycling signage that made invisible costs tangible.
AI's environmental footprint is in roughly the same condition sustainability was in twenty years ago: real, measurable, and largely invisible to the people generating it.
The talking cat video you like, your new campaign html coded in Claude, your AI assistant organising your messy folders — each carries a cost in water, energy, and compute that the user almost never sees or is able to quantify most of the time.
There is a role for Communications in making that cost legible to stakeholders. The applications range from the tactical (making the footprint of a given AI task legible to the person generating it) to the systemic (education campaigns on mindful AI use).
The role runs first inside the organisation — helping employees understand the footprint of their tools — and then extends outward, into how the public learns to think about AI consumption.
Consumption is also not simple or equal. The environmental impact of the same AI compute varies sharply by where it runs: a data centre in a region powered largely by hydroelectric energy carries a different footprint than one running on coal. Whether that variance gets disclosed, offset, or hidden is itself a reputation question or a crisis waiting to happen.
3. Token psychology
Tokenomics — the commercial economics of AI compute — is where the industry conversation currently sits. I would pivot it to token psychology: what unlimited (not technically unlimited and increasingly hardened by providers’ pricing policies, but let’s use this for the sake of example) token access does to the user, not the provider. Also referred to as cognitive offloading or cognitive surrender.
The early signals suggest a familiar pattern on the horizon — a technology that begins as unconstrained convenience and ends with the user negotiating their own limits.
Soon we might be looking at AI detox centres where people are going back to analogue, and screen-time caps for children extending into token caps on AI use.
We need to think hard about what we use AI for and in what capacity. I think the mechanism worth borrowing from elsewhere in tech is the SLAs.
AI providers sign service-level agreements with their clients — commitments on uptime, performance, response time. I wonder what an SLA signed in the other direction would look like. So I’m committing to it for my own usage. A company crafting it for its people’s usage. It might look like this:
What gets committed to: which tasks stay fully human (the craft-building ones), which are AI-assisted (human leads, AI supports specific subtasks like research, summarisation, or drafting fragments), which are AI-led (AI produces the output, human reviews and signs off).
What gets measured: frequency of unassisted work in defined categories. Cognitive performance over time to see the impact of AI usage in our thinking, reasoning, and execution.
Who holds the commitment: the individual, the team, the function head — probably all three at different levels.
Token budgets as the concrete artefact: a quantified ceiling rather than a vague principle. An employee with a weekly token budget must decide how to spend it. That forces intentionality.
Even when I write this, I am conscious of how quickly it can be misread, so I will unpack:
At the personal level, this lands closer to a gym membership or a self-imposed screen-time cap than to surveillance. You, as the user, are setting the ceiling for self-protection, and your data stays with you. Voluntary, self-directed, designed to preserve a capability you want to keep.
At the organisational level, the framing has to be sharper. An employer-set token budget for a junior employee is surveillance-adjacent unless the framing is explicit: this is a protected category of unassisted work, not a productivity cap. Although soon organisations might lean that way to optimise token costs (post for another time). The closest parallel is continuous professional development (CPD) policy: a structural commitment that prioritises long-term capability over short-term output.
That is the register to write it in. Designed wrong, it reads as Big Brother watching you. Designed right, it reads as the organisation taking responsibility for the (cognitive) wellbeing of its people.
Communications can be the driving force in shaping that conversation — again starting inside organisations first and then extending to how the public comes to think about cognitive surrender/offloading in an AI-saturated environment.
4. Ethics by design
Across the three areas above, there is a single binding principle: ethics must be designed into AI systems, not retrofitted after deployment.
AI for PR articulates this clearly. It is a shift from treating ethics as a patch to treating it as a fundamental building block of system design. The principle is broader than reputational risk. It covers what gets built, how it gets built, and what gets constrained — and it applies to any function involved in AI design, not Communications alone.
But it has a specific consequence for Communications. This is the principle underneath the Trust consequence dimension I unpack in the Two Clocks Framework. If ethics has to be designed in, the function that reads reputational consequence has to be in the room at conception — not called in after deployment fails.
That position is only credible if the function has the capabilities to occupy it. Two Clocks names three, with more likely to emerge as the technology evolves:
AI literacy — not technical depth, but sufficient understanding to engage governance conversations and challenge system design decisions.
Reputational risk architecture — mapping where AI systems create trust exposure across decisions, surfaces, and agentic deployments, before deployment rather than after failure.
Coherence judgement — looking across everything the organisation says about itself in the surfaces AI systems draw on, and reading whether those parts are telling one story or several.
The direct output of these capabilities, exercised in time, is a seat at the AI governance table — so that the four areas in this newsletter are designed in, not retrofitted after failure. For more on Communications’ role in AI governance, see this previous newsletter.
The slow thinker role does not stop at the organisation's edge. The question is whether Communications recognises the scale of the room it now stands in — and the consequences of not speaking up.