From the newsletter archive
Do this before connecting AI to your email marketing
Litmus The State of Email 2026 Report opens by talking about the way AI is changing the inbox.
Hubspot State of Newsletter Growth 2025 report has a dedicated section on AI use in newsletter creation.
The data is clear: using AI in your email marketing…
Contributes to higher ROI
- 63% of marketers see a positive ROI from using AI in email marketing and
newsletters (Hubspot) - 75% more likely to achieve ROIs above 45:1, likely because high adopters approach email differently (Litmus)
Speeds up email creation
- 79% of marketers agree that AI helps them spend less time on manual tasks
- It now takes 1-3 days for teams to create one email, vs two weeks in 2024 and up to one week in 2025 (Litmus)
Do the boring and tedious jobs so you don’t have to, like personalisation, segmentation, testing, and analysis.
If you’re feeling excited to implement AI in your event email marketing straight away, hold on.
It’s not as simple as connecting the AI tool and asking it to do your campaigns for you.
Your whole system needs to be ready for it.
Hubspot ran an experiment last year to personalise its content using AI and increased conversions by 82%.
But the first test did not go as planned, and so they refined, tested, and refined again.
We refined the AI’s training data, improving its ability to interpret user behavior and guess the job-to-be-done. We tested. We iterated. And, after months of tweaking and adjusting, we finally hit the sweet spot.
And to be ready for AI, you need to make sure it is as tidy and as organised as possible:
- Clean up your tags/groups/segments
- Create a systematic naming convention for all your emails, campaigns, groups, and tags
- Record your email creation workflow so that you can ask AI for help to improve it
If everything lives in your head and random notes here and there, AI cannot help optimise your email marketing.
Examples:
Email marketing data hygiene needs to level up from human-centric to one that the AI can understand
A human marketer compensates for messy data with judgement. An agent does not. It will confidently do the wrong thing at scale.
1. Naming and structure
Humans: “VIP” as a tag is fine. You know who it is referring to, and when looking at the list of names, you know who’s who.
Agents: “VIP” is meaningless without a definition that the agent can read. You need either:
A naming convention encoded as metadata (segment_type: speaker, tier: 1)
Or a separate property the agent can filter on cleanly: VIP-speaker-may2026; VIP-sponsor-tier1-feb2025; etc.
2. Consent has to be machine-readable
Humans: Verbal consent or one in a DM somewhere lives in someone’s head.
Agents: Every contact needs explicit, readable properties:
- consent_marketing_email: true/false
- consent_timestamp: 2026-02-14T19:32:00Z
- consent_legal_basis: “consent” / “soft_opt_in” / “legitimate_interest”
3. Lifecycle stage as data, not vibes
Humans: You know Sarah is a “previous attendee who’s gone quiet”. You’d write her a personal line.
Agents: Sarah needs a stage property that the agent can pivot on:
- lifecycle_stage: prospect / waitlist / ticket_holder / attended / lapsed
- last_event_attended: “February Bash 2026”
- attendance_count: 2
Without it, the agent treats Sarah identically to a cold lead.
4. Identity resolution
Humans: You spot that “sarah@gmail.com” and “sarah@mybusiness.com” are the same person.
Agents: They’re two contacts unless you’ve merged them or set a shared identifier, and they risk getting the same message twice, or worse, different messages entirely.
You and I both know that you don’t have this level of organisation in your email marketing today. Possibly not even in your CRM.
But without stringent data hygiene that AI can understand without making assumptions, using AI in email marketing is pointless.
You’ll just be going faster nowhere.
This is a piece from my newsletter archive, sign up here to get emails like this in your inbox first.