Why your AI strategy isn’t delivering ROI and how to fix it

Everyone’s talking about how they’re using AI. Salespeople summarize calls, marketers write emails, product teams brainstorm new features. It feels like the way work gets done is changing. But ask how AI fits into their actual workday and most will say, “I copy my notes into ChatGPT, get a response and paste it back into my doc.”

That’s not transformation — that’s translation. The digital equivalent of printing an email just to read it. The winners won’t be the best prompters. They’ll be the ones designing systems that think with them.

Your AI ROI strategy isn’t working — here’s why

McKinsey research shows that nearly eight in 10 companies report using generative AI — yet just as many report no significant bottom-line impact. Take a deep breath and let that sink in — what you’re doing today isn’t working. 

The potential for genAI in GTM teams is massive, but the chatbot shortcut delays that value. It replaces system design with surface-level Q&A, deferring true transformation. There are two core disconnects in the AI-ROI conversation:

• CEOs assume AI’s value lies in efficiency through headcount reduction.
• GTM teams chase the wrong tools or fail to reimagine the right applications that deliver ROI.

CMOs often lean into the efficiency mandate, but effectiveness is what drives value. Your CEO needs both. Here’s how to ground that approach:

• Better insights (effectiveness): AI compresses the distance between knowledge and decision, enabling teams to pivot faster, catch patterns earlier and capture opportunities others miss.
• Process optimization (efficiency): AI enforces winning workflows, eliminates variation and reduces error rates — improving conversion while freeing up strategic capacity.

CEO translation: AI ROI isn’t about labor savings alone. It’s about systems that are faster, smarter and structurally aligned to growth. Simplify your message into two levers: efficiency and effectiveness.

Dig deeper: Marketers report surging ROI as genAI moves from pilot to practice

The duality of AI strategies

Efficiency optimizes process. Effectiveness scales intelligence. Chatbots do neither. They’re useful for ad hoc Q&A, but disconnected from GTM execution. Rather than fragmenting teams with ad hoc chatbots or digital twins that seem impressive but deliver little in cost savings or performance, focus on tangible ROI — cost savings through efficiency and performance lift through effectiveness.

• The right idea: Automate operational workloads to reduce manual effort and improve precision, speed and consistency in execution.
• The right tool: GenAI (not chatbots) configured with business rules and rich knowledge that streamline processes and enhance decision quality.

GenAI as process optimization: Efficiency strategy

Look for ways to automate operational tasks to enhance speed and productivity. How many reports take an ops person a full day each week to compile? How many campaign reports does your team build each month? These are necessary tasks, but they consume a massive amount of time — and when someone’s out sick or on vacation, progress stalls.

As a CMO, I once demanded this discipline. It was state-of-the-art. Now it’s obsolete. This is where AI must go to work. Operational roles centered on routine processes should be the CMO’s first target. It may sound abrupt, but these roles will be nearly eliminated within two to three years. If you haven’t deployed projects in this area, you’re already behind.

While writing this, I’m also building a 30-step AI workflow that reverse-engineers a company’s GTM strategy from its website — mapping audiences, capabilities, messaging and more. It runs on roughly 1,000 lines of Python and 200+ lines of JavaScript, using NLP, NER and entity clustering to extract, validate and prioritize insights. That’s possible today.

How did I get here? I followed one piece of advice: use ChatGPT to draft a functional spec, then deploy it in a low-code automation tool, piece by piece. The chatbot’s broad, general knowledge is perfect for this use case.

CEO translation: Reduces operational costs, shortens decision lag time and increases confidence in data-driven decisions.

GenAI as knowledge infrastructure: Effectiveness strategy

You start with AI for speed, then something goes wrong (i.e., a hallucination, a wrong answer or a generic suggestion that ignores your business reality). That’s normal — but it’s also a signal.

This is the moment you realize fast isn’t the goal. You need correct. You need context. And you need it all to scale with precision. That’s when you pivot — from assistant to strategist, from velocity to veracity.

How does this play out in practice? Think about how often teams recreate the same deck, campaign or messaging sequence because the last version is buried in someone’s folder. Sales uses one insight, marketing another, product teams yet another. And too often, we reach for the easy button, pitching the wrong ideas to a prospect or anchoring a new piece of content on outdated assumptions.

That’s the effectiveness gap — when organizational insight is siloed and unevenly distributed across critical GTM teams. It only worsens as complexity grows. Multiple products, industries, personas and geographies make the permutations of what “right” looks like nearly infinite.

This is where genAI can truly shine, although few teams have yet to realize it. The knowledge of an LLM is generic and broad — but your GTM strategy is narrow and deep. To unlock effectiveness, you must replace the LLM’s generic knowledge with your own GTM strategy.

Just as radiologists use AI trained on millions of medical images — not generic chatbots — to spot tumors and anomalies, GTM leaders need AI trained on their strategy, messaging and customer insights to create true effectiveness.

AI as IP for your GTM strategy

Your GTM knowledge isn’t just content — it’s intellectual property. As AI commoditizes the value of broad knowledge, expert-trained AI models become your IP moat — your key lever to drive value and ROI. Treat this knowledge like a product: curated, maintained, versioned and deployed across your organization.

When you do, AI becomes a real-time copilot — not a chatbot, but a knowledge engine that adapts to strategic applications and bridges the gap between client and prospect needs and the GTM elements they need to understand. This is more akin to library science than computer science — a gap that sits between technical expertise and business knowledge.

For more than two years, I’ve built expert-trained LLMs that turn organizational knowledge into institutional assets for sales enablement and content creation. Often spanning over 20,000 rows of code, these models replace an LLM’s broad, generic understanding with the narrow, deep intelligence of a company’s GTM strategy. The result: no prompt engineering, no hallucinations. Teams simply talk to the LLM as they would a colleague.

Dig deeper: The hard truth about what AI will do to GTM

Building your knowledge infrastructure

This isn’t ChatGPT reading a PDF. It’s the infrastructure layer of your GTM. Proprietary knowledge is contextualized and surfaced the moment your teams need it. You can start now. Gather your core GTM strategy elements and place them in a shared location. 

Common assets include:

• Objectives
  • What do you want from your GTM investments?
• Messaging and positioning
  • Value proposition communication: Clear articulation of the unique value your solution delivers to target audiences.
  • Market positioning: How to position against competitors, emphasizing unique features and approaches.
• Capabilities and differentiation
  • Product capabilities highlighting: Focus on key functionalities that stand out.
  • Competitive analysis: Comparative insights that show where your solutions excel.
• Personas and their challenges
  • Need-based solutioning: Address pain points and connect needs to value and capabilities.
  • Value propositions: Proven ways to express solutions for each persona.
  • Segment-specific targeting: Tailor messages to the unique challenges of each role.
  • Lifecycle stages: Identify how needs evolve across the customer journey.
• Examples of successful content
  • Format and writing style: Voice and tone that reflect your brand.
  • Context: Audience understanding and real-world application of your strategy.

The most effective approach is to feed this content into a vector store (semantic database) and point your LLM to that source. I use the OpenAI Assistant infrastructure, which will transition to the Responses API by mid-2026. 

Even a simple setup using File Search creates a knowledge source that’s 30-40% stronger than a generic chatbot. While your goal should be a 90% improvement, don’t let perfection delay progress. Higher-quality data improves adoption and trust — although it does require active knowledge management.

Your objectives in this exercise is to have:

• The right idea: Codify expertise and make it usable everywhere.
• The right tool: Proprietary GTM knowledge embedded in AI systems.
• The right outcome: Faster strategy cycles, sharper personalization, higher conversion and a scalable execution engine.

Looking ahead, in three to five years, this knowledge infrastructure will become standard in B2B commercial applications. It’s a portable asset — one that connects across workflows, CRM and MAP systems — accessible to all. If you haven’t started down this path, you’re already falling behind.

CEO translation: This isn’t about cutting headcount — it’s about higher signal quality, faster insight cycles and deeper GTM leverage.

The AI loop: Process and knowledge converge

Process drives speed. Knowledge drives relevance. Together, they redefine what’s possible in GTM. Whether you start with process or knowledge, the two inevitably converge.

Automate the basics — formatting, reporting, data hygiene. Let strategic gaps guide knowledge investments, and let those investments fuel more advanced automation. Strong processes create better insights, and smarter insights power smarter automation.

Dig deeper: How AI flipped the funnel and made GTM tactics obsolete