Why AI Agents Need Better Briefs Than Humans Do

Why the instruction that was good enough for a human hire can quietly cost you when you hand it to an agent

The tension in briefing marketing work has always been there. What is different now is where those briefs – and the budgets attached to them – are increasingly going. As more marketing decisions and execution are handed over to AI agents, the quality of the instruction becomes much more important.

A brief written for a person was often only partially complete. That was usually fine because the person receiving it brought their own understanding of the business to the task. They knew the customers, understood the commercial priorities and could recognise when the literal instruction wasn’t quite the same as the intended outcome.

An agent doesn’t have that background unless you give it to them.

That means defining what success looks like is not simply a useful addition to the brief. It is what allows the agent to make the right decisions in the first place.

The Brief Was Always Half-Finished

A lot of the most important information in a marketing brief has historically been left unsaid because experienced people were expected to fill in the gaps themselves.

Ask a media buyer to reduce cost per lead, for example, and a good buyer won’t simply pursue the cheapest leads available. They will consider lead quality, the sales team’s ability to convert those leads and the commercial value of different customer types. They may know that a £40 lead is considerably more valuable than a £20 lead in certain circumstances, even if that distinction never appeared in the original instruction.

That understanding comes from experience and context. It develops through sitting in sales meetings, looking at CRM data, speaking to clients and understanding how marketing performance translates into revenue.

The brief therefore never needed to contain everything. Some of the thinking was carried by the person doing the work.

That becomes a problem when the person is replaced, or increasingly supported, by an agent that does not have that same context. If the commercial assumptions are not made explicit, the agent has no reliable way of knowing which ones matter.

What An Agent Does With The Same Brief

Consider the instruction to generate more leads. Given that objective, an agent can potentially find and process far more opportunities than a human could. But if the brief doesn’t explain what makes one lead more valuable than another, there is nothing stopping it from pursuing the cheapest sources of volume.

The result can look successful on the surface. Lead numbers increase, cost per lead falls and the campaign appears to be moving in the right direction. Meanwhile, the sales team may be seeing little improvement in the quality of the pipeline.

The agent hasn’t necessarily made a mistake. It has followed the objective it was given. The problem is that an assumption which a human marketer might have made automatically was never actually included in the instruction.

The same issue applies to paid media more broadly. If an agent is asked to optimise performance but is primarily given platform metrics to work towards, it may naturally focus on the things the platform can measure and reward most easily: clicks, conversions, engagement or efficiency metrics.

Those numbers can all be useful signals, but they are not necessarily the same thing as business success. Unless the agent understands what the business ultimately values, it can become extremely effective at optimising towards the wrong outcome.

A Good Brief Makes Context Explicit

The answer isn’t simply to make briefs longer. An agent doesn’t need pages of instructions covering every possible scenario. It needs the information that an experienced marketer would otherwise infer.

That means being clear about which customers matter and why, what the business is trying to achieve, which outcomes are commercially valuable and which metrics are simply indicators of progress.

It also means defining what a good lead actually looks like.

For a lead generation business, the objective might appear to be increasing lead volume while reducing cost per lead. But the more useful instruction might be to increase the number of sales-qualified opportunities within a particular cost range, while protecting conversion rates further down the funnel.

That distinction is obvious once it is written down. It is much less obvious to an agent working from a campaign objective and a set of platform metrics.

The same principle applies across paid media. If an agent is going to make decisions about budgets, audiences, creative, products or bidding, it needs to understand the commercial context within which those decisions should be made. Otherwise, it will optimise according to the signals available to it rather than the outcomes the business actually cares about.

Why Briefing Is Now A Commercial Skill, Not An Admin One

This changes the role of the brief. It is no longer just an administrative document that gets the work started. It becomes a way of transferring commercial judgement into an automated system.

That makes briefing a much more valuable skill than it might appear.

Someone who can write a good brief for an agent needs to understand the business well enough to recognise the information that the agent is missing. They need to distinguish between a useful performance signal and the actual commercial objective, understand the trade-offs involved and decide which constraints should shape the agent’s decisions.

That is a different skill from simply knowing how to use an AI platform.

As more execution becomes automated, technical knowledge of individual platforms will still matter, but commercial judgement becomes increasingly important. Someone needs to decide what the system should be trying to achieve, what it should take into account and where it should not be allowed to optimise without additional context.

Someone also needs to own that decision-making.

If the brief is where commercial intent enters an agentic workflow, then responsibility for the brief is effectively responsibility for how that intent gets translated into execution.

The Brief Becomes Part Of The Work

For a long time, marketers have tended to think of the brief as something that happens before the real work begins. You gather the requirements, explain the objective and hand it over to the person responsible for execution.

Agentic AI changes that relationship.

The more capable the agent becomes, the more important the quality of the instruction becomes, because increasingly there is less human intervention between the brief and the outcome. The system may be capable of analysing the data, making decisions, executing changes and learning from the results, but it still needs to understand what the business is actually trying to achieve.

You would not ask a talented marketer who had just joined your business, knew nothing about your customers and had never seen your sales data to start making decisions without giving them that context. The fact that an AI agent can execute the work faster does not remove the need to provide it.

In fact, it makes that context more important.

The shift towards agentic AI therefore isn’t just changing how marketers execute campaigns. It is changing where some of the most important thinking needs to happen.

The better agents become at execution, the more carefully marketers need to think about the instructions they give them.