How a social media agency assesses staff skills

What to take away

Plan hiring around oversight and interpretation, not production. Tools now draft copy, cut clips and build reports, so the scarce skill is judging output and explaining it to clients.
Treat the UK GDPR and PECR as part of the data-law baseline. The Data (Use and Access) Act 2025 amends parts of that framework; check ICO guidance as requirements change.
Score candidates against a written rubric before you interview, or you will hire for confidence rather than capability.
Review the forecast quarterly, because the tools and rules do not sit still.

The skills shifting fastest

From making content to checking it

Generation is cheap now. Judgement is not. A planner who can read a model's draft, spot a claim that would breach the Advertising Standards Authority's (ASA) CAP Code, and rewrite it is worth more than one who produces volume. They should also check influencer disclosures, usage rights and whether a post is suitable for its platform.

Making vs Checking Content

Old emphasis

Core task: Produce volume
Value: Generation
Client ask: Supply output
AI role: First-draft tool

New emphasis

Core task: Check and edit
Value: Judgement
Client ask: Explain safety
AI role: Agent-style supervision

Tools such as ChatGPT, Google Gemini and Canva Magic Write can speed up drafting; Looker Studio can support reporting. For a social-media agency, test whether a candidate can review an AI-assisted caption, short-form video or report for unsupported claims, disclosure gaps, rights issues and misleading performance summaries before it reaches a client.

Data handling as a day-one requirement

Campaign work touches personal data constantly: audience lists, comment moderation, influencer contracts and lead forms. The Data (Use and Access) Act 2025 amends parts of the UK GDPR, the Data Protection Act 2018 and the Privacy and Electronic Communications Regulations 2003 (PECR); it does not replace them. The Information Commissioner's Office (ICO) is the UK's data protection regulator. Candidates should know when to consult its guidance and how to document their handling decisions.

Turn those requirements into an interview test: ask candidates how they would limit access to audience lists, handle moderation and lead-form data, and record the consent decisions relevant to a campaign.

Building the forecast

Score the roles you actually need

A forecast is only useful if it changes a decision. List the tasks for the next four quarters, then score each role against the rubric below. Anything below 12 goes on a development plan, not a job description.

Role Scoring Rubric

AI output oversightx3
Data and consent handlingx3
Paid media literacyx2
Client explanationx2
Tool change adaptationx1

Weight

AI output oversight: x3
Data and consent handling: x3
Paid media literacy: x2
Client explanation: x2
Tool change adaptation: x1

Score 1

AI output oversight: Accepts drafts unedited
Data and consent handling: Knows the basics
Paid media literacy: Reads dashboards
Client explanation: Reports numbers
Tool change adaptation: Waits for training

Score 5

AI output oversight: Documents checks and edits
Data and consent handling: Applies rules to live campaigns
Paid media literacy: Explains budget shifts to clients
Client explanation: Connects numbers to decisions
Tool change adaptation: Tests and documents changes

Multiply each score by its weight and total. A weighted score of 12 or more suggests the person can carry the role now; below that, plan training. For platform skills, development routes include Meta Blueprint, TikTok Academy, LinkedIn Marketing Labs and Google Skillshop. Treat course completion as evidence of study, not proof of campaign judgement.

Make the rubric practical with a role-specific exercise: ask the candidate to review an Instagram or TikTok post and an ad in Meta Ads Manager, then explain the human sign-off, rights and disclosure checks, and any audience-data consent evidence they would record.

Watch the paid media release cycle

Paid-social work may span Instagram, TikTok, LinkedIn and YouTube. The relevant tools include Meta Ads Manager, TikTok Ads Manager and LinkedIn Campaign Manager; YouTube campaigns can be managed through Google Ads. Someone should own the release notes for these tools and turn relevant changes into a monthly note for account managers.

Give one person responsibility for the updates and a slot in the team meeting. Each note should say what changed, which campaigns or accounts may be affected, who must act, and what guidance the account team should give the client.

When to update the forecast

Fixed review triggers

Set dates rather than waiting for discomfort. Review the forecast when a client sector starts requiring consent records, a platform changes its advertising tools, or a hire changes the team shape. A quarter without a review is itself a trigger.

Each trigger needs a named owner and a one-page output. A review with no written output gets repeated next quarter.

What to measure

Track three numbers. First, time from tool change to documented guidance. Second, share of outputs with a recorded human check. Third, client retention where the agency explained a rules change before the client asked. None of these needs new software.

If the numbers drift, the forecast is wrong, not the team. Rewrite the rubric weights for the next hiring round.

Common questions

Is this forecast specific to England?

The data framework described here applies across the UK. This article focuses on agency hiring and client work in England; agencies elsewhere in the UK should account for their own client mix and any sector-specific requirements.

How often should we rerun the scoring rubric?

Quarterly is enough for most teams. Rerun sooner if a platform changes its advertising tools or a client asks for consent evidence you cannot produce.

Do we need a data specialist on staff?

Not always. One trained account lead who can apply the rules to live campaigns often covers a small agency. Larger teams handling sensitive sectors should consider dedicated support.

Does AI oversight belong in every role?

Yes, at a basic level. Anyone publishing client work should be able to say what they checked before it went live.