Email marketing is the channel with the highest ROI in digital marketing, but also the most saturated. The average professional’s inbox receives between 100and 150emails a day. Standing out in that context no longer depends on sending more: it depends on sending better, with more relevance for each recipient at the most appropriate time for them.
AI transforms email marketing in three dimensions: content personalisation, prediction of the optimal send time, and intelligent automation of flows.
Personalisation 1:1 with generative AI
Basic personalisation (inserting the contact’s name) improves the open rate by between 5% and 10%. Real personalisation — adapting the content, suggested products, examples used and the call to action to the recipient’s specific profile — can improve it by 40-100%.
With generative AI in the email flow, the process works like this: the system receives the contact’s data (sector, interaction history, position in the funnel, recent web behaviour), passes it to the language model with personalisation instructions, and generates the email body in real time for that specific contact. It is not a template with variables: it is a text written specifically for that person.
The result doesn’t sound like AI when the system is properly configured with the brand’s voice guide. It sounds as though someone took the time to write it with that specific recipient in mind. Because technically, they did.
The predictive model for optimal send time
Email opening behaviour data shows clear patterns: some people open emails on their mobile at 7:30 before getting up, others check them all at 11:00 after their first meeting, and others only open professional ones on Tuesday and Thursday afternoons.
A Bayesian model trained on each contact’s open history can predict with good accuracy when that contact is most likely to open the next email. Rather than sending to the entire list on Tuesday at 10:00 (because someone once decided that was the best time), each contact receives their email in their own optimal window.
The impact on open rate is consistent: between a 15% and a 30% improvement compared to fixed-time sends for most B2B lists.
Subject lines with AI: beyond "Test this vs that"
Language models are particularly good at generating subject line variants because they were exposed during training to enormous quantities of marketing text, including subject lines with their open rates. The model has an implicit knowledge of which patterns work.
The correct methodology: use AI to generate between ten and twenty subject line variants for each email, categorised by angle type (urgency, curiosity, direct benefit, personalisation, number, question). Select three strong candidates from different angles for the test. The A/B test with three variants across 30% of the list determines the winner, which is sent to the remaining 70%.
This process can increase the open rate by between 20% and 40% compared to using the first subject line the team comes up with.
Automating complex flows with decision AI
Traditional email automation flows are static: if they do X, they go to path A; if they don’t do X, they go to path B. Flows with decision AI are dynamic: the system evaluates multiple signals simultaneously and decides what the most appropriate next communication is for that contact at that moment.
Example: a contact who has opened three consecutive emails, visited the pricing page twice and downloaded a case study from the same sector should not receive the next generic nurturing email. The system detects those signals, automatically activates the accelerated conversion sequence and notifies the sales rep.
To see how we implement predictive email at BAI, see our Email Predictivo product.