Only a year ago, the industry was debating whether brands should hire prompt engineers. It already feels quaint. Prompt engineering has been overtaken by something far more powerful: agentic systems. Allow me to digress for a moment.
Almost every evening, I create bedtime stories with ChatGPT, and my children give me ideas based on their imagination. Currently, we’re deep in an episodic saga involving the family, Star Wars, Sonic and Hogwarts. But the story arc has become formulaic and repetitive. Even my kids are tired of beating Darth Vader with a Sonic Spin-Dash every time. My appreciation of these stories has diminished.
I realised two things very quickly.
Firstly, I value the output less.
Behavioural economists call it the effort heuristic: we equate value with visible labour. The harder something appears to be, the more we trust it. When a strategy, an insight, or a concept is produced quickly, even if it’s brilliant, we instinctively treat it with suspicion.
A 2023 Nature Scientific Reports study found that identical artworks labelled “AI-made” were rated up to 62 per cent less valuable than when labelled “human-made”. The work didn’t change. Only our perception did.
This bias follows AI into the workplace. We celebrate AI for automating targeting, analytics and optimisation. But the moment it enters the world of ideas, writing, ideation or planning, we hesitate. Not because outputs are weaker, but because the effort is less visible. It’s a psychological barrier, not a technological one.
Secondly, I need to put more thought into the concept to derive meaning from the output.
Creativity isn’t just about generation. It’s about intention.
From ad-hoc prompts to agentic capability
Early prompt libraries at PHD were a practical bridge. They helped our people structure thinking, access information consistently and reduce hallucinations. Prompts sat inside Studio Workflow in Omni, guiding AI through the steps of campaign planning: hypothesis generation, research, and refinement.
But prompts were always a stopgap.
The real breakthrough came when we moved beyond isolated prompts to agentic systems. These are AI processes that think in multi-step sequences.
Agentic capability means the model forms hypotheses, extracts keywords, runs desk research, checks source credibility, rejects weak ideas while strengthening good ones, synthesises evidenced conclusions and generates new hypotheses.
This is why the conversation about prompt engineers now feels outdated. Not because prompts don’t matter, but because agentic systems can generate, adapt and sequence their own prompts internally. The real value sits somewhere else entirely.
Agentic systems don’t improve outputs linearly; they compound them. They deliver depth, consistency and speed at a scale humans simply can’t replicate alone.
From prompts to hypotheses: where human creativity enters the system
This shift isn’t just technical. It fundamentally changes how value is created inside agencies.
Concept engineering is the art of embedding human intent, judgement, creativity and problem-framing at the front of the process, giving agentic AI the right ‘soul’ to work from.
AI can scale and refine thinking, but it can’t originate purpose. It can extend a narrative but won’t work out why the narrative matters. It can evaluate hypotheses, but not choose which tension is worth exploring.
That’s human territory.
Hypotheses provide the ‘why’ while Agentic processes deliver the ‘how’. This is the moment when AI stops replacing labour and starts expanding human ingenuity. It’s the part of the story that is genuinely exciting and, yet almost entirely under-recognised.
Why the value gap is a good thing
Agentic systems are already delivering a step-change in speed, quality and depth, but many people still aren’t appreciating it.
That isn’t a problem; it’s an opportunity.
The organisations that close this psychological value gap, who help people trust, use and feel the power of agentic intelligence, will be the ones who capture disproportionate benefit. This is exactly where we at PHD now sit: ahead of the capability curve, and ahead of the adoption curve.
The joy of rediscovering thinking itself
What surprises me most about agentic systems isn’t the speed or the scale. It’s something far smaller and more human, the rediscovery of what it feels like to think freely.
The bedtime-story problem showed me the trap: when AI generates without intention, the output becomes predictable, and the joy dissolves. But the moment I bring a real concept, a tension, a twist, a purpose, the model gives me the freedom to explore it without the drudgery of mechanics. The thinking becomes lighter. The ideas get bigger. The process feels playful again.
And perhaps that’s the real message here. For us, Agentic AI is creating an army of thinkers, which gives everyone the opportunity to develop strategy and add value.
By Amit Patel, Executive Director, PHD UAE.
