I work out where AI pays off in your CRM and marketing, and what to build first. I've run CRM from the inside for fifteen-plus years and built an AI email tool on Braze.
What this AI marketing audit is
“AI marketing audit” can also mean using AI to audit your marketing. This page is about the other kind. I go through your CRM and marketing, work out where AI would pay off, and rank what to build first.
It runs in the same shape every time. Every candidate gets the five questions below, and the audit ends in a ranked build list.
The feature nobody used
At Tibber I built Spark, an AI email and content tool, in-house and integrated with Braze. It did three things: brief-to-send email, SEO content, and an AI tone-of-voice scorer. The whole platform was designed around the scorer.
Email was the runaway winner. SEO content came second. The tone scorer went unused, because the copy was already good enough that nobody checked it.
The rule I took from that is the one this audit runs on. Build for the job people actually hire the tool for.
The AI in CRM use cases I weigh
The usual candidates, all but the last in marketing automation:
- Copy generation, from subject lines to brief-to-send email
- Send-time and channel choice
- Propensity and churn scoring
- Segmentation
- Content for SEO
I’ve built and shipped three of these: brief-to-send email and SEO content in Spark, and propensity scoring on Databricks, XGBoost and MLflow. They get the same questions as the rest:
- What job would it do, and for whom?
- Is the data there, in Braze or wherever your events live, with consent to use it this way?
- Where would it run: a trigger, a segment, a step in a customer journey you already have?
- How would you know it helped? A holdout, measured against the same journey without it.
- Who keeps it running after the build?
The ranked build list
The output is a ranked build list, short enough to act on Monday. The top line is what I’d build first. Spark’s tone scorer would have sat near the bottom of mine.
If the top line is a program you want built, that’s a lifecycle marketing program, with the scope agreed before we start. If you need someone inside the team to own CRM, list included, that’s a Fractional Head of CRM.
An AI CRM consultant who has run CRM
In AI for CRM, most of the questions on that list are about data, consent, triggers and holdouts. I’ve worked on those from the inside, as the person the numbers land on: Head of CRM at IKEA Ingka Group across 30+ markets, and at Bauer Media Group across 20+ brands in six European markets. Spark came out of Tibber, energy-tech in four markets. Its receipt is below.
I build with Braze, Databricks, XGBoost, MLflow, Python and Cloudflare, hands on keyboard. For a small working example of propensity scoring, this site scores you as you read it, this page included: p(book), a logistic model with hand-set weights over 12 signals, shown signal by signal on the home page. It’s the toy version of what runs on Braze, Databricks, XGBoost and MLflow.
Based in Malmö, a short train ride from Copenhagen. The audit runs in English or Swedish, for teams across Sweden, Denmark and Norway.
andersson