Meaningful Engagement and AI
Meaningful Engagement and AI
I set out to face the AI fear head-on — testing whether a model could help project developers engage meaningfully, and where the human still has to hold the line.
The explosion of AI has a lot of us asking existential questions about whether our skills are still needed — and if our jobs cease to exist, what will be left for us to do. Lately, I decided to face the AI fear and started digging into the specifics of how various AI tools actually work. By systematically unpacking the potential for AI to replace human effort on a granular level across a broad-spectrum ESG role, I have banished some of that existential dread.
01The hypothesis
By focusing on how I can use AI to combine what makes me and my experience unique with the potential to work more efficiently, I set about testing the following hypothesis with Claude: that AI can play a significant role in helping project developers engage meaningfully — but that meaningful engagement is only possible through human connection.
My first step was to ask Claude if it could help me write a thought-leadership piece on the use of AI for community engagement and stakeholder management. As well as several questions designed to support me on my publication journey, I was most interested in the last thing it asked — about how I wanted it to contribute:
This taught me my first lesson: I needed to be very careful about what I ask, so as not to instil bias. In the interests of anchoring the experiment without limiting imagination, I chose a framework from the European Centre for Not-for-Profit Law, which identifies three elements of meaningful engagement: Shared Purpose, Trustworthy Process and Visible Impact. I tabulated where I saw the potential and the weaknesses across the three elements, then asked Claude to do the same — and combined both into one table.
02Underestimating the limits
Claude essentially mirrored my thoughts on its ability to create administrative efficiencies. It also acknowledged the substantive points: that meaningful engagement requires integrity and a commitment to share decision-making power, and that there is no substitute for relationships — in particular the ability to sit, listen, and change one's mind.
The extent of Claude's technological limitations surprised me. It is very self-aware about its weaknesses — its inherent bias, from how it was trained; the way it can operate opaquely, giving the impression of objectivity when it can actually be used to rationalise decisions already made. It also has a tendency towards hallucinations and misrepresentations that could lead to nuance being missed or comments being invented. The probabilistic way it makes predictions makes it utterly unsuitable for assuming community needs or requests.
Claude highlighted the positive potential to open opportunities for audio/visual formats and real-time translation. But it made no reference to the reliance on — and costs of — power, connectivity and physical facilities, which can be serious limitations in remote areas. It did teach me about the need for consent in terms of data and AI use, which I hadn't yet taken on board as a novice.
03The practitioner's work
I'm as quick to admit to taking comfort from the fact that Claude cannot make business decisions or build relationships, as Claude is to confess to being unreliable and biased. But it boils down to us rebuilding our understanding of how it works and how it can support and enhance our work, rather than replace us. AI can enable a shift away from the transactional elements of recording data and managing engagement cadences — towards the more interesting, challenging work of consensus-building and influencing decision-makers.
04Floor, not ceiling
This exercise has given me comfort that there is space for humans and AI to collaborate to improve engagement outcomes. And I fully agree with Claude's concluding comment:
AI can substantially raise the floor on the mechanics of engagement — reach, accessibility, recall, consistency — but it cannot raise the ceiling on its integrity. Integrity still requires binding mechanisms — FPIC, enforceable agreements, independent grievance redress — and a genuine willingness to change course.
— Claude, on the limits of its own roleUltimately, however, the mere fact that it categorises trust-building as "relational work" tells me it's not mature enough to fully grasp human interaction in all its nuance and unpredictability.
Audrey Hackett
Co-founder of Solenne Solutions, a boutique ESG and strategic communications advisory. Audrey works with project developers on community engagement, stakeholder management and meaningful, defensible ESG practice.
