Every business I speak to is planning AI into 2027. New tools, new budgets, new promises.
Here is the catch. Everyone is buying from the same shelf. The model you use is the model your competitor uses. So the tool is not where the advantage sits.
The advantage sits in what your AI knows: about your buyer, about where they are in the journey, and about what your team should do next. AI is only as good as the context you give it. In most businesses that context is scattered, stale or missing.
AI has gone furthest where it needs to know the least
The State of Martech 2026 report surveyed 208 marketing and marketing operations leaders across around 70 AI use cases. The pattern is clear.
Content and experience use cases sit at 89% AI adoption. Copy ideation is close to universal: only 14 of 163 respondents are not using AI for it. Commerce and sales sits at 49%.
The report explains why in a single line: “You can’t score a lead in a standalone AI tool that doesn’t know your pipeline. You can ideate a content strategy with nothing but a prompt and some brand context.”
Writing a post needs a prompt. Deciding which buyer to call this week, and what to say, needs to know the buyer. That is where the revenue is, and it is exactly where AI struggles without a connected foundation underneath it.
The report has a name for fixing this: context engineering. Assembling the right data, rules and instructions so AI can act intelligently in a specific situation. Scott Brinker, who has tracked martech since 2008, has called context martech’s word of the year for 2026.
Everyone builds the roof first
Most AI plans start at the top. Pick the tool, run the pilot, wait for results. But AI is the roof. Underneath it sit the two layers that decide whether it holds: your data and tools, and the orchestration layer that connects them.
In most businesses that foundation is a Frankenstack: a marketing and sales setup stitched together from tools that were never designed to work as one. Website behaviour in one place, email in another, LinkedIn in a third, call notes in someone’s head.
Put AI on top of that and you are building on a cracked foundation. It does not fix the cracks. It runs on them, faster.
The report gives a B2B example that will feel familiar. A buyer spends four minutes on the enterprise pricing page, then raises a support ticket. Three systems hold three signals about one buyer, and the account team never sees them. As the report puts it: “The question isn’t, ‘Do you have the data?’ You probably do somewhere.”
What your AI needs to know about your buyer
Look at a typical CRM record. Name, email, company, maybe a job title. That is a one-dimensional record. It tells AI who to email. It does not tell it what to say, when, or why.
A three-dimensional record answers three questions:
- Who is the buyer: role, company, fit, and what they care about
- Where are they in the journey: what they have read, clicked, downloaded and asked
- What will help them take the next step: the right content and the right conversation, at the right time
Nobody on the team types that record in. The system builds it, stage by stage:
- Capture: every lead lands in one place, with its source
- Profile: persona and fit are added automatically
- Engage: behaviour is tracked, clicks, visits and downloads, not just opens
- Nurture: buyers who are not ready get relevant content until they are
- Convert: sales sees everything marketing knows, and moves when the signals say so
Get that right and AI has something real to work with. Skip it and you are asking AI to guess.
What it looks like running
We built this for ourselves first: one connected system, with two AI-powered layers on top.
At the front, AI-powered lead generation. LinkedIn outreach runs through Dux-Soup and Expandi, joined to the CRM by an n8n engine. When someone connects and starts a conversation, the engine enriches their details, loads them into the CRM, profiles them and suggests the right nurture. Every new lead arrives with context, not just a name.
In the pipeline, AI opportunity plans. For every live opportunity, the engine reads the full record: stage, behaviour, content consumed, conversations and nurture history. It checks that against our sales playbook and email library, then writes a plan for the next 30 days.
Here is the difference on one illustrative record, anonymised:
| |
One-dimensional record |
Three-dimensional record |
| Who |
Sales Director, SaaS business |
Sales Director at an 80-person, PE-backed SaaS firm. The MD wants a forecastable pipeline before the next funding round |
| Journey |
New lead |
Took the Frankenstack Quiz, downloaded a guide, attended a webinar, clicked six nurture emails, visited pricing three times, booked a call |
| Next step |
Follow up |
Send a pipeline predictability model this week. Set up a joint session with the MD. Push for the close once both are engaged |
The plan takes around 10 minutes to generate. Reading and validating it takes longer, and that is deliberate. A person approves every action and every email before anything goes out. What comes back is a clear set of instructions: what to do, when, and what moves the opportunity forward. Every plan is refreshed monthly.
We are now piloting a learning loop, so what worked and what did not feeds into the next round of plans.
The 2027 test
Before you sign off another AI budget, ask three questions:
- Can you see a buyer’s behaviour in one place? If it sits across four tools, your AI sees a quarter of it
- Is your sales process written down? Stages, rules, what good looks like. AI can only follow a playbook that exists
- Who approves what your AI produces? Speed without review is how the wrong message reaches the right buyer
If any answer is no, start with the foundation. The AI will still be there, and it will work far harder for you.
See AI running on a connected system
Book a demo and we will walk you through the whole journey on one live system, from first LinkedIn connection to an AI opportunity plan naming the next move. No jargon, no pressure. Or start with a 30 minute call.
Book a demo Book a 30 min call
Recent results include a £20m qualified pipeline for an IT outsourcing business and £1.5m of new revenue for a nationally recognised restaurant group within 12 months.
Related reading