Audience research for AI marketing tools: fix the brief

What to take away

What is the most common mistake when scoping an AI marketing tools target audience?

Most teams begin with capability. They read what a model can predict, then hunt for people who match. The audience becomes a by-product of the software.

That order hides the real question. A target audience exists to change something: the message you send, the channel you fund, the leads you pass to sales. If no decision moves, the segment has no job.

A blunt test helps: name the person who will act differently, and what they will do. Vague answers on both counts mean the brief isn't ready.

Write the decision at the top of the brief, then leave it there when the segment list changes. The AI marketing tools audience research guide for England sets out how method, evidence and legal checks fit together.

Which questions should you settle before you open the tool?

Then write the boundary in testable terms. Marketing managers at firms with 50 to 200 staff that bought a CRM in the past 18 months can be checked. Ambitious SMEs cannot.

Firm details such as headcount or a recent purchase are usually visible in your own records. Intent, budget and authority usually are not, so agree how you will judge them.

Worked example: a vendor of AI-assisted lead-scoring tools could test marketing managers at firms with 50 to 200 staff that bought a CRM in the past 18 months. A candidate segment might use Salesforce CRM and be considering Salesforce Einstein; another might use HubSpot Marketing Hub's AI features for campaign personalisation. Check firm fit in CRM records and campaign outcomes in Google Analytics 4, where available, then interview candidates to find out whether the workflow would change a real decision.

Email and text marketing are governed by the Privacy and Electronic Communications (EC Directive) Regulations 2003 (PECR); the ICO's Guide to the Privacy and Electronic Communications Regulations (PECR) and Direct marketing guidance explain consent and the soft opt-in. For cookies and similar technologies, use the ICO's Guidance on the use of cookies and similar technologies. Personal data processing and profiling fall under the UK General Data Protection Regulation (UK GDPR) and Data Protection Act 2018; for AI-related processing, consult the ICO's Guidance on AI and data protection. Ask your data protection officer (DPO), where you have one, to review the proposed data and purpose.

How do you score candidate audience segments?

Use one rubric for every candidate, and score before you discuss preferences.

Score each row from 1 to 5, multiply by the weight and add the totals. Two segments within five points of each other are too close to split on numbers alone. Pick the one you can evidence more cheaply.

Score with the person who owns the budget in the room. A rubric agreed on paper tends to drift once preferences surface.

Where does regulation fit into audience research?

Digital rules are increasingly written in coordination rather than in isolation. The Digital Regulation Cooperation Forum workplan shows how the main regulators align their programmes, which signals where scrutiny is heading for profiling and automated decisions.

Ask early whether an inferred attribute can be explained to the person affected. If the answer is no, cut the attribute from the brief rather than defending it later.

Can you research an audience with a small sample?

Yes, if the questions are narrow. Six focused conversations often beat a survey of hundreds, because you can follow a thread and hear the reasoning behind an answer.

Recruit for variety: different company sizes, seniority levels and tool experience levels, then compare notes against your written segment rules and record where they disagree.

Write each conversation up on the day, while detail is fresh. For an AI marketing tools audience, ask how the person currently segments customers, which tools and workflow they use, who owns the buying decision, what the AI feature would change, and what evidence or data would justify adopting it. Record answers against the segment rules, including objections and buying triggers, rather than treating interest as purchase intent.

Common questions

How many audience segments should we keep?

Two or three for a first pass. Any more and the team cannot hold the differences in mind, so messaging drifts back to averages.

How often should we revisit the definition?

Review it when a decision changes, not on a calendar. A pricing move, a new channel or a launch is a better trigger than a quarterly slot.

What if our data cannot support the segment we want?

Narrow the segment until it can. Start with what your CRM holds, add interviews, and leave the ambitious model for a later round.

Do we need consent to score existing customers?

It depends on the purpose and the lawful basis you rely on. Marketing by email or text usually needs consent or the soft opt-in, so check the guidance before you build the list.