AI-Powered Campaign Assistant

The problem

Our client was delivering several campaigns every year, for different sectors, products and audience types, and through multiple channels.

Their challenge was threefold:

1. Knowledge: nobody had a clear idea which tactics or assets had worked for a previous similar campaign, so every new campaign was a "best guess".
2. Silos: different teams owned different channels, and there was no coordination of performance or learnings against benchmarks.
3. Reporting: monthly reporting required a day of someone's time to pull all the data together into one place, by which time it was already out of date.

Whilst they had reviewed several "dashboard" platforms on the market, our initial consultation identified that this was really a time and institutional knowledge issue, and that a dashboard would only simplify an existing problem.

What the client needed more than anything else was a campaign data analyst to manage the process for them.

What we built

We began by auditing all available data across their web, email, social, CRM and paid ads platforms. This allowed us to identify data quality gaps, format inconsistencies (for example dates and KPIs), establish governance and ownership, and determine which tasks were consuming the most team time.

We then created an index to ingest and normalise the data into a single search index, flattening it into a single, consistent data set that shared common values across various campaign facets. This meant we could match values such as "5 Dec 2025" versus "05/12/25" and "CTR %" versus "Click-through %".

Custom rules around benchmarking, audiences, campaign theme and seasonality were key, so we created bespoke business logic to enable clearer data filtering, provide a mechanism to flag under- or over-performance at any time for any channel, and link campaign acquisition to performance and drop-off further down the website funnel.

Finally, we applied an LLM model on top of the data, which would allow a marketing team member to query the campaign knowledgebase in natural language, and added automated email and Slack alerts if a live campaign was underperforming or if a campaign was set up with missing facet data.

The finished product was a bespoke AI solution that:

1. Opened up three years' worth of multichannel campaign data insights to anybody within the marketing team at the push of a button.
2. Allowed campaign issues to be addressed promptly to reduce wasted spend by constantly measuring against benchmarks.
3. Provided insights into which assets and messaging were resonating most effectively for a particular audience and theme, so that future campaigns were built on evidence, not gut feeling.
4. Flagged acquisition-to-on-site-funnel drop-out issues to enable experimentation to maximise campaign spend ROI.
5. Became a real AI data analyst capable of condensing days of work into just a few seconds.