My organisation is asking me to use AI to reduce budgets. What should I do?

14 Sept

Communications teams are being asked to use AI to cut budgets, but this delivers weak returns, ignores the work AI is creating, and misses the strongest investment case the function has had in years.

2027 planning guidance is asking the external affairs and corporate communications function to use AI efficiency savings to reduce headcount and agency costs.

It’s an argument as old as the first industrial revolution: a new technology arrives and management expects it to pay for itself and improve margin through productivity savings. The natural reaction of any public relations leader is to protect their budget and resources.

But this is a more fundamental issue and the arguments are in flux. In this newsletter I’ve attempted to set out the evidence that Ben Verinder and I collated for AI for Public Relations: A How-To Guide for Implementation and Management, along with studies published since the book went to print.

These caveats aside, I’ve identified five counter arguments.

Acknowledge AI savings and make the case for change management
Headcount is the wrong metric to capture the value of AI
AI is generating new work that we are only just beginning to understand
The reputational impact of AI as a stakeholder must fall within the public relations remit
Use AI to recover and rebuild capacity

If there’s anything that I’ve missed please let me know in the comments and I’ll add an update.

But before we explore the arguments I want to share some personal news. I submitted my PhD disseration this week which explores the conditions necessary for public relations to be recongised as a management function.

Thank you for your support over the past six years. There’s an Appendix that explores how I’ve used LinkedIn and Substack to share ideas and gather feedback. I’m planning to turn it into a paper in time, once I’ve defended the disseration.

Acknowledge AI savings and make the case for change management

We have to be pragmatic and acknowledge that we’ve been using generative AI for management and communication tasks for the last two years. The evidence is in plain sight in coverage reports, first drafts and posts on intranets and social media.

A credible defence must start with conceding this reality. Arguments about relationships, judgement and nuance lose the room.

Ther counter-argument is that communication teams are at different stages of AI adoption within their workflow. This is a culture change programme and needs to be properly planned and funded.

The argument: AI will change the make-up of the function, but it is too early to say how. AI adoption requires investment in workflow, training and socialisation.

Headcount is the wrong metric to capture the value of AI

A Gartner survey of 350 managers in $1bn revenue companies found no correlation between the scale of headcount reduction and return on AI investment. Organisations getting return from AI were building workflows, training teams, and creating new “amplification” roles to support adoption.

A second study of AI layoffs reported that employee trust is the single most important predictor of return on AI investment. It found that cutting staff to free capital for AI produced the worst outcomes of all.

There’s a shiny new toy aspect to AI. Managers overestimate gains ahead of actual evidence. A Federal Reserve Bank of Atlanta survey of more than 750 managers called out this paradox.

The argument: Headcount reduction ahead of measurable AI gains leads to weak return on investment at best and may be damaging at worst.

AI is generating new work that we are only just beginning to understand

The working assumption is that AI reduces workload within the communication function. This view has held up for almost two years. But it has become clear that second-order AI effects are generating more work, both externally and internally.

The inbound problem shows up as the use of AI by the public. A letter of complaint or freedom of information request takes seconds to draft and costs nothing. A paper due to be presented at the AAAI Conference on AI Ethics and Society in October calls this phenomenon agentic flooding.

The Information Commissioner’s Office published advice in May for public organisations dealing with higher volumes of requests, greater complexity, and requests that misstate the law and need clarification before they can be processed. The BBC reported in August that complaints that previously ran to a single page of A4 are now typically more than 20 pages.

The internal problem is the requirement to deliberately redesign workflow to retain human oversight. This is the issue that the AI for PR community addresses throughout the book. AI creates additional work in editing, rewriting, and verification.

Ben Verinder wrote this week about the problem of trying to edit a document produced by a poorly trained AI model to make it sound more human. He said the job is upstream, in the governance, training and the socialisation that sets out how AI is used in a function.

The argument: AI is adding to workload through a flood of machine-generated inbound correspondence on one side, and the need for human oversight on the other.

The reputational impact of AI as a stakeholder must fall within the public relations remit

Reputation within AI models, so-called Generative Engine Optimisation (GEO), is our strongest investment argument. When a stakeholder asks an AI tool about an individual or organisation, the answer is parsed from whatever the model can find. This is typically earned and owned content.

A study by Muck Rack analysed more than a million links and found that between 84% and 89% cited earned media. A smaller study by Hard Numbers and Onclusive of the world’s 100 largest brands put the number at 61%.

The results are highly contextual by topic and market but the directional signal is clear. The content that AI models use to serve answers is the work of the communication function. But here’s the issue: no function yet owns the stakeholder perspective of AI. It falls between customer service, marketing, sales and communications.

Your plan should include a line to benchmark GEO visibility and build the infrastructure across the organisation to manage AI as a stakeholder.

The argument: AI models answer questions from content that the communication function produces. Claim ownership of that relationship by benchmarking GEO visibility and building the capability to manage AI as a stakeholder.

Use AI to recover and rebuild capacity

Every technology wave in my working life has been pitched by vendors as a productivity tool but has delivered a heavier load to the communication function.

Email in the 1990s shortened the cycle time of work. The web shortened it further in the 2000s and did away with the news cycle. Mobile in the 2010s, and messaging devices in the 2020s have collapsed it completely, removing the boundary between life and work.

The impact on practitioners is well documented. CIPR and PRCA data published in 2024 reported that nine in ten practitioners had experienced poor mental health in the past 12 months.

AI is different from the technologies that came before in that it offers the opportunity to remove work but that needs to be carefully and deliberately planned.

The management question is what happens to the time saved. It is both a cultural and economic issue. Unplanned, the pattern of the last 30 years will continue. Alternatively the released capacity can be invested in the team itself.

The argument: AI is the first communication technology in 30 years that could eliminate work rather than add it, but only if management deliberately decides how to reinvest the time.

So, what’s the plan?

The AI savings argument will be made whatever you say. The more strategic question, and the one I’ve tried to answer, is the opportunity that AI provides to rethink the function and its value.

Organisations shifting budgets to fund AI are getting worse returns than those reinvesting. AI is adding to the workload of the function and creating a new remit that has yet to be owned. And for the first time in a generation a technology is releasing capacity that the function could invest in its own people.

The practical step is to start measuring. Keep a simple record of every task AI assisted with, the hours it released, where those hours were reinvested, and what the reinvestment produced. Also track the new work created by AI.

Golin Ketchum’s Jeff Beringer sets out this approach in a chapter he wrote with Jonny Bentwood in AI for PR. Evidence provides a defensible account of value creation, he argues. And by the time next year’s planning guidance arrives you’ll have your own data and arguments.

And i you need help making the case for your budget for 2027 please get in touch.

Have a good week ahead.