Like most marketers, I thought AI was going to change everything overnight. Write my copy. Generate my images. Brainstorm campaign ideas. Optimize performance. It felt like a cheat code for marketing. And to be fair, it can do all of those things. It’s made it easier to get started, explore ideas, and punch above your weight, especially if you’re on a smaller team.
But two years in and I'll be honest: the reality is a bit more nuanced.
The problem isn’t whether AI can produce outputs. It’s whether those outputs actually hold up when you’re trying to run real campaigns, under real constraints, with little time to spare. Because when you’re on a lean team, “almost there” is still quite far.
Where AI helps (and where it falls short).
Let me walk through where AI has actually landed for me and most of the growth marketers I talk to.
Across the board, AI is excellent at getting you started. Need email copy? You’re out of the blank page instantly. Need creative? You’ve got options in seconds. Need campaign ideas? It’ll give you a solid list to react to.
That alone is a big deal. For small teams especially, removing that initial friction is half the battle. But getting started and getting something out the door isn’t quite the same thing. There’s usually a gap between the first output and something you feel confident about actually using, whether it’s tone, fit, or just a sense that it’s not quite there yet. So, you make a few tweaks. Then a few more. Then you pause and wonder if you’re helping the tool, or if the tool is helping you.
The real constraint: Context.
The issue isn’t that AI isn’t powerful enough. It’s that it doesn’t have enough context.
We had a customer recently share their experience trying an AI feature on a non-Adobe marketing platform. They gave it a detailed prompt: define a segment of users on a free plan, last active between 30 and 60 days ago, with a product engagement score above 70. All the variables were there. The output was, in their words, nonsensical — it returned a segment that included recently active paid users and excluded most of the free-tier accounts they were trying to reach. Their reaction was, “I’ll never use this again.”
That’s the risk with AI in marketing. The first experience matters. If the output isn’t trustworthy, adoption stops before it starts. AI is intelligent, but it’s not omniscient. The difference between a helpful output and a useless one is the same as the difference between asking a random person for advice versus asking a coworker who knows your business inside and out.
Marketing is an inherently context-heavy discipline. The old adage — the right message, to the right person, at the right time — is fundamentally a data and context problem. Personalization isn't a tactic anymore; it's the baseline expectation, and executing it requires knowing your customer data, business logic, past performance, and brand inside and out. That's a lot of context. Too much to paste into a chat window. For most mid-market teams, there isn’t a budget to waste on broad, generic campaigns. Precision matters. Context isn’t just helpful — it’s the job.
Introducing Project Halo: Marketing at the speed of thought.
When we started building Project Halo, the goal wasn’t just to add AI into existing workflows. It was to rethink the workflow entirely.
So, we asked: what if the traditional point-and-click interface of martech solutions disappeared?
What if, instead of clicking through menus and configuring rules, marketers could simply describe what they want to achieve? And what if the system actually had enough context to produce outputs they could trust?
Project Halo is our answer. It’s a specialized AI solution designed for lean teams to generate entire marketing campaigns using customer data, business context, past campaigns, and brand assets — all from a single prompt.
What you will do with Project Halo.
Project Halo will be the first AI solution created specifically for growth marketers that uses a prompt to generate entire campaigns using your data, your brand context, and your historical performance.
We're essentially building marketing-native LLM, one that isn't reasoning from general knowledge but from the specific context of your business — your customer segments, your past campaigns, your brand assets, your offers, your journeys. This is vibe marketing: the ability to move from intent to execution at the speed of a conversation, without sacrificing the intelligence and personalization that actually drives results.
For marketers on lean teams who need AI to deliver confident outputs today, here's what we're building:
- Generate and refine complete campaign plans: From objective to audience to content and timing, in a single conversation.
- Instant campaign performance answers: Ask natural language questions about what's working, what isn't, and why.
- Audience curation from your data: Upload customer lists, define objectives, and get smart segment recommendations grounded in actual behavior.
- Automated brand kit generation: Project Halo will read your website, analyze your existing assets, and generate a brand context layer that informs every output.
What this looks like in practice.
The clearest way to show the difference is through a real campaign scenario. Consider a free-to-paid conversion campaign, the kind every B2B SaaS marketer knows well, and knows it is painful to build properly.
In the past, you'd build a rule like this:
IF customer_segment = "free_users"
AND subscription_expiry = "May 2026"
AND engagement_score ≥ 80
THEN enroll in journey = "paid_conversion_v3"
WITH offer = "first_year_discount_20pct"
That rule took someone an afternoon to build properly. They had to figure out the right engagement score threshold (is 80, right? Maybe it should be 70?), identify an existing, high-performing journey for this use case, find and validate the offer code, test the segment before launch, and manually QA the trigger logic.
With Project Halo, we want you to simply describe your goal: "I want to convert free users who are most likely to pay before their trial ends in May. Use our best-performing offer."
The system would then analyze your audiences, historical conversion rates by engagement score, available journeys, and active offers. Using all of this context, Project Halo will build the campaign, recommends the threshold based on past performance data, and flags anything that needs your review.
Go from workflows to momentum with customer engagement designed for growth teams, powered by AI.
We believe this conversational way of working represents a shift in how work gets done. It makes sophisticated marketing strategies easier to execute and iterate on — even for small organizations. Campaigns like welcome series, win-backs, product education flows, and product announcements can be built and refined in minutes instead of days.
For growth marketers navigating complex stacks and increasing performance pressure, it’s not just another feature. It’s a new way to work. Project Halo is currently in private beta with a select group of customers who are actively helping us build this the right way. They are stress-testing it with real data, live campaigns, and business constraints, with a full launch coming soon.
Alex Erickson leads industry-focused Product Marketing for the Adobe Experience Cloud, specifically across High-Tech, Financial Services, and Manufacturing industries. In his eight years at Adobe, Alex has held various roles focused on M&A, business strategy, and product strategy. Alex comes from a diverse background, including social entrepreneurship, management consulting, investment management, and government.
