Back-to-Front AI Adoption
Apply AI to existing customers before expanding into acquisition.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 4
- Confidence
- 97%
This framework prioritizes AI adoption by moving backward through the customer journey: delight, then engage, then attract. Teams first apply AI where they possess the richest proprietary context—support histories, product usage, onboarding needs, and renewal signals. Once an experiment demonstrably improves the experience of existing customers, the underlying capability can move upstream into personalized outreach and sales engagement. Only after those applications work should the company emphasize acquisition and content generation. The sequence reduces speculative experimentation, anchors investment in customer outcomes, and creates a learning loop based on real interactions rather than public data alone. It also helps teams distinguish meaningful AI integration from adopting tools simply because they are novel.
Origin
Extracted from Marketing Against The Grain during Kieran Flanagan and Kipp Bodnar's discussion of where marketers should begin integrating AI across the go-to-market journey.
Core principles
- 01Evaluate AI through the customer journey.
- 02Start where proprietary customer data creates the strongest advantage.
- 03Improve outcomes for existing customers before pursuing acquisition volume.
- 04Expand only after measuring the preceding stage.
How to run it
- 1
Map the Customer Journey
Describe how customers move through attract, engage, and delight. List the consequential interactions, data, and desired outcomes at each stage.
Pro tip Use customer outcomes rather than internal departments to define the journey.
Watch out Do not begin by selecting an AI tool before identifying the interaction it should improve.
- 2
Start With Delight
Apply AI to customers who already use the product, beginning with support, onboarding, ongoing usage, expansion, or renewal. Use proprietary product and customer information to make the experience more useful and available.
Pro tip Support is a strong starting point because answers can be evaluated for accuracy, speed, and resolution.
Watch out Keep human escalation available when the model lacks confidence or the issue is sensitive.
- 3
Move Into Engage
Reuse validated AI capabilities to personalize consideration-stage interactions such as sales outreach, recommendations, and meeting preparation. Combine customer context with accurate product information.
Pro tip Trigger outreach after a meaningful behavior rather than generating unsolicited messages at scale.
Watch out Personalization without relevance can feel invasive and damage trust.
- 4
Expand Into Attract
Apply proven capabilities to research, content creation, localization, landing pages, and acquisition. Measure whether AI facilitates the next action rather than distracting visitors.
Pro tip Test narrowly against an existing baseline before redesigning the entire acquisition experience.
Watch out A conversational interface can reduce conversion if it encourages exploration instead of action.
In the wild
A software company connects its support automation to approved documentation and account context. The AI answers routine questions around the clock and transfers uncertain or complex cases to a human agent.
→ Customers receive faster answers while support staff spend more time on exceptional cases.
After proving customer-facing AI responses, a company uses the same product knowledge to draft outreach when a prospect completes a meaningful action. The message recommends the most relevant package and explains why it fits the prospect's role and company.
→ The outreach becomes more contextual than conventional name-and-company personalization.
A marketing team tests AI-generated landing-page variations only after establishing evaluation and governance through support and engagement projects. It compares each variation against a static control and monitors both conversion and lead quality.
→ The team learns whether upstream personalization creates incremental value without risking the entire website.
Common mistakes
Starting With Content Volume
Content generation is accessible, but beginning there can prioritize output over meaningful customer outcomes and proprietary advantages.
Automating the Whole Journey at Once
A broad rollout makes it difficult to isolate errors, understand impact, or transfer learning between stages.
Optimizing Engagement Instead of Outcomes
Longer AI conversations may look engaging while delaying the specific action the customer needs to take.
Is it for you?
Best for
It is best for established companies with existing customers, support data, and measurable go-to-market workflows.
Not ideal for
It is less useful for pre-customer startups without meaningful support, usage, or retention data.
From the transcript
“Always look through the lens of the customer journey.”
“So when I say back to front, I would start with like the delight stage like customers who are already... The people who are already…”
“Then I would move into engage, and then I would move into attract.”
From the episode
OpenAI’s ‘Foundry’ Release Will Change Marketing Forever...