Why Most AI Roadmaps Are Backwards
How Pricing Units, Not Features, Should Drive Product Strategy
In my prior post, “How to Monetize Your AI Roadmap”, I laid out a simple but uncomfortable reality: most AI features never become revenue.
That post introduced the four-level ladder: Capability, Feature, Product, and Digital Labor. It explained where revenue actually shows up and why most teams stall at the bottom.
This post picks up where that framework leaves off. And it’s the question I get most often after walking teams through the ladder. Because once you understand where revenue lives, the next question is obvious:
Why do so many AI roadmaps still fail to reach it?
The answer is sequencing. Most AI roadmaps are built backwards because they follow traditional SaaS approaches that no longer make sense given the massive economic differences.
The Backwards Roadmap Fallacy
The dominant AI roadmap pattern looks like this:
Identify model capabilities such as summarize, generate, or chat
Turn those capabilities into features
Ship quickly to drive usage
Ask Sales and Pricing to figure out monetization later
This pattern made sense in traditional SaaS, where marginal costs were close to zero, and bundling features rarely hurt economics.
In AI, it breaks immediately. Inference, infrastructure, and support costs show up from day one. Expectations are set early. Once customers internalize AI as bundled functionality, charging later feels like a price increase rather than new value.
As I argued in “How to Monetize Your AI Roadmap”, if pricing is not explicit, it is implicitly zero. The roadmap may look full. The revenue path usually isn’t.
The Wrapper Trap: Jasper’s Rise and Fall
Jasper is a clear illustration of what happens when roadmaps optimize for features rather than pricing units. It reached unicorn status by offering “AI for marketing copy.” Early growth was real, and adoption was strong.
The problem was economic design.
Jasper’s roadmap focused on templates, brand voice, and UX polish. But its pricing unit never moved beyond the underlying capability - easier access to the model.
As large language models improved, marketing copy became a commodity feature inside the platforms where marketers already worked. HubSpot, Salesforce, Google Workspace, and others embedded similar capabilities directly into their core products.
Because Jasper never climbed beyond Level 2 on the ladder, “Feature”, its pricing power collapsed as soon as the capability became free elsewhere.
This is the wrapper trap. If your roadmap stops at features, your business is exposed to commoditization the moment the platform catches up.
Why the Four-Level Ladder Is a Roadmap Problem
In the prior post, I argued that most AI initiatives stall between Level 1 and Level 2.
What this post adds is why.
Roadmaps built around capabilities and features naturally anchor teams at the bottom of the ladder. They optimize for shipping, not for ownership.
To reach Level 3 or Level 4, “Product” or “Digital Labor”, the roadmap must be designed around a pricing unit first, instead of the model or feature.
The Better Sequence: Price First, Build Second
Roadmaps that successfully monetize AI reverse the order.
Define the pricing unit: Who pays? For what outcome? From which budget?
Design the workflow: What level of reliability, trust, and integration is required to justify that price?
Build the features: Only what is necessary to support the economic model.
In this model, the roadmap is not a list of features. It is an economic document.
This is the practical application of the ladder. Pricing power is designed at inception, not discovered later.
Level 3 in Practice: Harvey and Vertical Ownership
Harvey is a strong example of a roadmap designed to reach Level 3.
They did not build a generic lawyer chatbot. They built an AI system embedded directly into the risk, research, and reputation workflows of elite law firms.
The roadmap was constrained by the outcome from the start.
Harvey integrates proprietary case law, firm-specific precedents, and high-stakes workflows tied to billable hours. Removing it would force firms to hire more associates to perform low-margin work.
That is why Harvey commands premium pricing. Not because the model is novel, but because the roadmap was designed to address a painful, expensive workflow.
That is Level 3 monetization by design.
Level 4 in Practice: EvenUp as Digital Labor
EvenUp shows what happens when a roadmap is built explicitly for Level 4. They do not sell software access. They sell work.
Their AI analyzes medical records and produces complete demand packages for personal injury law firms. Pricing is tied directly to outcomes, often on a per-document or share-of-recovered-value basis.
Because the AI finds revenue lawyers would otherwise miss, the pricing unit sits directly on the customer’s revenue line.
This is the top of the ladder: “Digital Labor”.
The roadmap did not start with features. It started with the question: what work can we reliably own and monetize?
M&A Implications: Why AI Roadmaps Collapse After Acquisition
The sequencing problem becomes even more acute in M&A. Most AI acquisitions are diligenced on technical capability and feature potential. Very few are evaluated on pricing compatibility or ladder position.
Post-close, acquirers often discover:
The AI was designed for bundling, not monetization
The pricing unit does not align with the parent company’s sales model
The roadmap assumed data accumulation was the ROI, not revenue
The result is predictable. The AI is absorbed as a feature. Monetization stalls. The deal thesis erodes quietly. This is not an integration failure. It is a roadmap failure that was visible in diligence.
Successful M&A Example
Databricks’ acquisition of MosaicML illustrates what happens when pricing alignment is explicit. Databricks did not buy Mosaic just for technology. It evaluated pricing compatibility. Mosaic sold efficiency. Faster and cheaper model training.
That fits cleanly into Databricks’ consumption-based pricing model, where customers already pay for compute and performance. The pricing unit matched. The roadmap survived post-close. Revenue synergy was immediate.
This is what climbing the ladder looks like in M&A.
A Diagnostic for Your Own Roadmap
Before committing real engineering time, ask four questions:
What is the pricing unit? Per user, per outcome, per unit of work, or unclear?
Who owns the outcome? Is there a specific role whose job improves measurably?
Does the math work at scale? Does revenue cover inference and infrastructure costs, not just in demos
What breaks if this is removed? Convenience, or revenue and productivity?
If these answers are vague, the roadmap is backwards.
Here are two additional examples that come to mind in CX:
HubSpot’s Content Assistant has an unclear pricing unit and is bundled. Usage is high. Incremental revenue is minimal.
Adobe Firefly launched with explicit generative credit pricing tied to usage. Adoption and revenue scale together.
Same era. Similar technology. Different sequencing.
Given the economics involved, AI roadmaps are no longer just product artifacts but substantive economic commitments.
Your roadmap is either a path up the ladder toward pricing power or an expensive exercise in building features customers expect for free. I’ve seen too many teams do everything right technically and still miss this. Choose the sequencing carefully.
If you’re building, buying, or operating in this space, I’d love to compare notes.
You can reach me at faraaz@inorganicedge.com or on LinkedIn.
Sources
AI Economics & Margins
Andreessen Horowitz – Navigating the High Cost of AI Compute — Link
Sequoia Capital – AI's $600B Question — Link
Bessemer Venture Partners – State of the Cloud 2024 (AI Gross Margins & NRR) — Link
Adoption vs. Revenue
McKinsey & Company – The State of AI in Early 2024: Gen AI adoption spikes and starts to generate value — Link
Gartner – Gartner Survey Finds 63% of Marketing Leaders Plan to Invest in GenAI — Link
Case Studies (Success & Failure)
EvenUp – Valuation & Outcome Pricing Model (TechCrunch) — Link
Harvey – Harvey AI Raises $100M at $1.5B Valuation (CNBC) — Link
Jasper – The Rise and Fall of Jasper AI (The Information) — Link
Adobe – Adobe Firefly Momentum & Run Rate — Link
M&A & Integration
