AI won’t transform your marketing if your workflow doesn’t

There’s a tension in the air. Many CMOs want AI to deliver a step-change in the next year or two. Yet most teams still run on processes built for manual craft. Bolt AI on at the end and you get faster copy, not a redesigned engine.

In regulated categories like pensions, the stakes are higher. You’re communicating long-term value and companies prioritise safety over move fast and break things. You have trustees, compliance, and Consumer Duty expectations to meet. If you don’t rewire how work flows, AI will stay a side-project.

The gap isn’t the model. It’s the operating system for creativity.

The typical path of creative : brief → concept → amends → traffic → QA → live

This assumes humans touch every micro-step. Useful for control; terrible for pace and learning. Forcing a human checkpoint on everything throttles impact.

The unlock comes when we treat AI agents as part of the team and redesign roles, guardrails and rhythm around them.

Three shifts that actually move the needle

1) Roles : Agents as junior teammates

Give agents a clear job description, an owner and a scope. In pensions, that might include: asset retrieval from an approved library, first-draft explainer copy for scheme members, content varianting for channels, and pre-flight compliance checks against a claims bank. Humans keep the high-risk calls: strategy, tone, final sign-off.

2)Guardrails : Governance you can audit

Map data access by tier (public, client-approved, restricted). Log prompts, outputs ,approvals and escalations. Capture why a decision changed. Especially if risk language was edited. In pensions, that audit trail is essential to show communications are clear, fair and not misleading.

3) Rhythm :From bursts to a continuous loop

Move from campaign bursts to an always-on cycle: sense → generate → test → learn →refine. Measure the loop, not just the launch. Track time to first draft, review touches per asset, and issues per 100 outputs. Let the loop improve the work week by week.

Thought experiment: the “Weekly Social Analyst” agent

Imagine an agent that watches your owned channels and relevant industry feeds, then delivers a Monday 09:00 note:

• Inputs (approved): last week’s posts and performance, a pensions claims library, tone rules, and a risk lexicon.
• Agent tasks (scoped): pattern-spotting, topic clusters, two or three draft post options, and a risk log with source links.
• Human gates (defined): strategist sets weekly priorities; compliance reviews any claim-bearing lines; design refines hero assets.
• Outputs (compact): a one-page plan, copy variants, flagged risks, and a simple test plan for mid-week and Friday.

The point isn’t the tool. It’s the shape of the work: agents handle the boring middle; people focus on judgement, taste and context. In pensions, that means more time explaining value clearly and less time chasing files.

What “good” looks like in pensions

• Claims discipline: performance or benefit statements link to an approved source; high-risk phrases are templated or blocked.
• Explainability: every material change says what changed and why alongside the asset.
• Separation of duties: the person who sets prompts isn’t the final approver.
• Kill-switches: if drift is detected—uncertain facts, tone issues—the agent hands back to a human.
• Model hygiene: data retention rules, red-team tests and prompt reviews sit in the weekly cadence.

The payoff!

Once you redesign the workflow, AI stops being a novelty. Speed and learning compound. Your team ships with more confidence and fewer handoffs. In pensions, that’s how you get clearer member communications without cutting corners.

The headline is simple: if you want AI-level outcomes, you need an AI-ready process. Start by redefining roles, guardrails and rhythm for a single journey. Let the loop do the lifting. Keep people where they matter most.

Source: AI-Orchestrated Creativity: Why Agencies Must Redesign by Chandni U

‍