Old habits die hard.
Businesses have been doing customer engagement wrong for decades.
Marketing teams build campaigns around products, not people. They segment audiences, design messages from first principles, choose channels, and deploy. It’s efficient, scalable, and fundamentally misaligned with the modern customer.
It also is hard to break away from as the problems are structural:
- Campaigns ask “which customers should see this message?” rather than “what does this customer actually need?”
- Personalisation is really just smaller broadcasts—the same offer, tweaked for a segment
- Success is measured in open rates and clicks, not genuine customer outcomes
- A skilled marketer’s judgment is the ceiling on relevance and speed
- Timelines stretch to eighteen months. By launch, the market has moved on.
Then Pega Customer Decision Hub changes everything.
Rather than asking marketers to predict customer needs, it asks AI to discover them in real time. Every interaction. Every signal. Every decision point re-evaluated by adaptive machine learning that optimises simultaneously for what the customer needs and what the business delivers. That’s the paradigm shift. And it only works if you design your implementations to let AI do what it does best.
Why the MLP Principle Works
An MLP—Minimum Loveable Product—isn’t a compromise. It’s the fastest route to proof.
The temptation is to go big: all channels, all data sources, all use cases. This delays learning by months, diffuses accountability, and kills momentum. By the time you prove value, you’ve spent a fortune and lost believers.
The right MLP is focused, fast, and evidence-driven. It chooses one channel. It uses a subset of data that already exists. It unleashes AI instead of writing rules. And critically, it proves that a fundamentally different approach to customer engagement actually works—with numbers.
The Five V’s provide the framework for getting this right.
The Five V’s: Your Blueprint for AI-Driven Customer Experience
| VISION | VALUE | VOLUME | VARIETY | VELOCITY |
|---|---|---|---|---|
| Customer-centric, not campaign-led | Measurable business impact | High-traffic, high-frequency engagement | Rich action library for true personalisation | Fast, incremental delivery—no big bang |
Each V is a design principle. Together, they define the conditions under which an MLP succeeds.
V1. Vision: Customer-Centric, Not Campaign-Led
Every decision in your MLP flows from one question: what is the best experience we can deliver to this specific customer, right now?
This sounds obvious. It’s not. Most organisations still think in segments and campaigns. An MLP with a weak Vision devolves into exactly that: the same twenty campaigns, now running inside CDH instead of Adobe or Unica. It’s not transformation. It’s just more expensive theatre.
A strong Vision for Next Best Action is this: every customer receives the most relevant, personalised engagement possible, at the moment it matters most, through the channel they prefer.
Why This Matters for AI
Adaptive machine learning thrives on clear, customer-centric objectives. The practical challenge in an MLP is that business-level outcomes—revenue, churn prevention, completion—can be slow to materialise or hard to isolate. So you’re often learning on engagement metrics: what customers click, what they respond to, how satisfied they are. That’s fine. The critical distinction is what you’re optimising for within those metrics.
The risk is subtle: confuse engagement metrics with campaign metrics. “Increase click-through on this offer” is campaign thinking. “Help customers discover what matters to them” is customer-centric thinking. The first makes the AI pick up on blast volume and promotional cadence. The second makes it learn what actually resonates.
So the question isn’t whether you have perfect business outcomes. It’s whether your objective is framed around customer discovery or around campaign performance. If your vision is clear—“customers should find the most relevant actions”—the AI learns that direction, even if you’re measuring it through engagement signals. If it’s fuzzy, the AI optimises for whatever proxy you’ve surfaced. And that’s often just the old paradigm in new clothes.
Here’s the practical test: Could you do this MLP with your existing campaign tool? If yes, the Vision hasn’t been properly applied.
V2. Value: Measurable Business Impact (The AI Multiplier Effect)
Value is where AI shows its teeth.
You choose an MLP use case where impact is clear and measurable. But here’s where most organisations underestimate AI: the value isn’t just in being smarter than campaigns. It’s in compounding.
A 2% improvement in relevance per interaction sounds modest. Multiply it across a million interactions per week, and you’re looking at significant revenue or cost impact. Now add AI’s continuous learning: each week, the model gets better. Without anyone manually tweaking it. The improvement compounds.
This is why you baseline metrics before go-live. Attribution matters. You need to prove AI is driving the result, not market conditions.
Value Levers AI Unlocks
- Conversion uplift: AI selects the right offer for the right customer at the right moment. Not based on a marketer’s guess. Based on hundreds of signals and real learning from thousands of prior interactions.
- Cost reduction: Proactive, AI-selected information prevents unnecessary calls and branch visits. Fewer inbound contacts. Same customer satisfaction. Better margins.
- Journey completion: Customers abandon halfway through onboarding, applications, setup. AI nudges them at the moment of highest likelihood they’ll complete. And it learns which nudge, which message, which offer works for whom.
Tying Capability to Value
The AI’s value only materialises if your use case leverages what makes CDH different: adaptive decisioning, real-time arbitration, cross-channel optimisation. If a simpler tool could deliver the same result, you’re not using AI’s real advantage.
V3. Volume: The Engine of AI Learning
Here’s what many people get wrong: a small-scale pilot teaches nothing about AI’s potential.
Adaptive models require data to become effective. Not just any data—real customer interactions, real outcomes, real feedback loops. A use case touching 10,000 customers monthly won’t generate the signal needed for the AI to learn well. A use case touching 10 million interactions monthly will.
Why Volume Compounds Value
This is the economic model of AI in customer experience. Small improvements in relevance, applied at massive volume, generate outsized results. And the model gets better every cycle. Week 1, it learns baseline patterns. Week 4, it’s discovered customer cohorts that humans never imagined. Week 12, it’s optimising long running journeys with carefully selected nudges at each stage.
Volume is what unlocks that curve.
Where to Start
For most organisations, the highest-volume touchpoints are digital: web and mobile. Authenticated sessions are gold—the system knows who the customer is, has context, can make genuinely personalised decisions.
Your MLP should reach a substantial proportion of your target customer base in that channel. If you go too narrow, the AI starves for learning data and the economics crumble.
V4. Variety: The Secret to True Personalisation
Personalisation is impossible with three stock campaign messages.
An NBA system needs a rich library of actions: offers, educational content, service nudges, recommendations. This gives AI real choices. And the more choices, the more likely the AI finds something genuinely relevant for each customer. Your moribund banners in the mobile app can become relevant again - and customers begin to learn they should take notice.
Yet many implementations populate the action library with the same campaigns they’ve always run. Now they’re just wrapped in an AI engine. Result: the AI picks between five offers, and every customer looks similar. Not 1:1. Not personalised. Just the old way, with an AI label on it.
Building the Action Library
Start with:
- Upsell and cross-sell: relevant products or services, tailored to what each customer likely needs based on their history and behaviour
- Educational content: guides, tips, explainers that help customers get more value from what they already have
- Service nudges: timely prompts to complete applications, update details, or take actions that benefit them
- Recommendations: tools, content, or resources tailored to their specific situation
The AI learns which action types work for whom, then optimises the library over time. But it needs sufficient raw material from day one.
The Variety Multiplier
Here’s what happens when your action library is truly rich: instead of optimising between “offer A or offer B”, the AI is optimising across dozens of actions and their variants. Suddenly, the percentage of customers receiving something relevant jumps dramatically. And the AI’s learning accelerates because it’s discovering patterns across more degrees of freedom.
V5. Velocity: Speed Unlocks Learning
Eighteen-month builds are the enemy of AI programmes.
By the time you go live, half your stakeholders have moved on. Sceptics have had eighteen months to dig in. And you’ve only just started learning from real customer data. That’s nine months late.
The MLP is built for speed—not by cutting quality, but by ruthless scope discipline.
Simplify the Technical Landscape
Single channel. Small but sufficient data set. Adaptive AI instead of complex rules.
Each additional channel multiplies integration work. Each additional data source multiplies engineering. Each additional rule adds configuration time and future maintenance burden. Your real constraint is not technology—it’s how fast you can learn from customers.
Let AI Do the Work
Here’s the power move: stop writing rules. Let adaptive models learn instead.
Rules take time to design and test. They encode last year’s assumptions. They need constant maintenance. Adaptive models begin learning from interaction one. They don’t require you to know in advance which customers want which messages—they discover that pattern in real time, from real data.
This makes them both faster to deploy and more effective than rules.
Build for Evidence, Not Perfection
The MLP should be the first sprint in an agile programme. Design it to answer:
- What engagement levels are realistic in this channel?
- Which action types generate the strongest customer response?
- What does the data reveal about customer behaviour that we didn’t know?
- What’s the true cost of delivery, and what does that mean for scaling?
These answers are worth more than any planning exercise. They’re the evidence that funds, shapes, and accelerates everything next.
What Success Actually Looks Like
At the end of a successful MLP, you should be able to say:
- AI is driving decisions. Customers are receiving personalised Next Best Actions, not campaign messages. And the data proves they’re responding better.
- Value is real and measurable. You can point to uplift you’ve created and attribute it to the new way of working.
- The AI is learning. Every week, the system improves without anyone manually tweaking it. The same patterns that humans would take months to spot, the AI discovers in days.
- Variety powers relevance. You’ve proven that a rich library of actions can be deployed at scale and that you understand which types resonate with which customers.
- You moved fast. Weeks, not months. And you have the evidence to fund and shape what comes next.
- The organisation believes. This is the cultural shift. The shift from “marketers decide” to “AI decides, guided by our values and objectives.” It sounds small. It’s everything.
That last point is perhaps most important. The shift from campaigns to Next Best Action is as much cultural as technical. The MLP is the moment that becomes real. It’s the proof point that changes minds, builds confidence, and creates the momentum for transformation.
Why This Matters Now
We’re at an inflection point in customer experience.
Organisations that still think in campaigns and segments are leaving money on the table and degrading the customer experience in the process. AI can do better. Dramatically better. But only if you design your implementation to let it learn.
The Five V’s—Vision, Value, Volume, Variety, and Velocity—are your roadmap. They ensure the MLP is customer-centric in purpose, measurable in impact, sufficient in scale, rich in options, and fast enough to build real momentum.
Apply them rigorously, and your MLP does something more important than proving the technology works. It proves that a different future is possible. One where every customer receives the most relevant, personalised engagement available, delivered by AI at the moment it matters most.
That’s the vision of Pega Customer Decision Hub. The Five V’s are how you make it real.