Agentic Content Engine Pilots
T&P runs several agentic content-engine pilots in parallel across accounts. The programme lines up on a similar transformation direction — how agentic pipelines could reshape content production — but specific needs and what was practical to pilot diverged sharply from one engagement to the next.
On Hawkstone the remit was the pilot slice itself: shaping the generation workflow end-to-end, deciding how data becomes localised social, video, and print for pubs across the UK, and keeping outputs honest when the ask swung between strategic and tactical speeds.
What shipped was pilot-grade: credible in review sessions and useful for deciding what to fund next — not a closed production handover. The wider programme carried on with new variants and follow-on pilots on other accounts.
Reference graph for the Hawkstone pilot — webhook-driven ingest, strategy, fan-out into copy and visual branches, then assembly into poster and final video. The wider programme shares a similar transformation direction; what was possible to build and demo still varied a lot by account — this topology reflects only Hawkstone in the pilot window.
Role
Led workflow architecture, n8n automation, cloud wiring for assets, and the integration surface producers used to run and inspect generations. Worked with strategy and production to agree what “good enough for this gate” meant for each drop, then iterated until the pilot outputs cleared those bars. Captured what worked, what broke under edge cases, and where the graph would need redesign if the work ever moved past pilot — so the programme did not have to relearn the same lessons from scratch.
Process
- Grew the workflow iteratively in n8n — wired stages as they became testable, then adjusted inputs, branching, and asset contracts from what failed or surprised in real runs rather than freezing the graph upfront
- Built n8n orchestration with routing, triggers, and enough logging to debug runs when timelines compressed
- Connected Azure and GCP for storage and retrieval, tightening how files moved between steps as the hand-offs became clear from testing
- Shortened the loop between prototype outputs and stakeholder review; trimmed scope when latency or quality risk spiked
- Closed the pilot with notes on limits, dependencies, and what a longer runway would require, while similar experiments continued on other accounts
Tech
n8n · Gemini · Azure · Google Cloud Platform · ElevenLabs · Veo