The more I work with AI—especially our AI-powered sales readiness tool, Shadow—the more I realize just how difficult it is to truly productize these applications. That’s not say other people, and us, won’t try. GenAI and large language models (LLMs) are vast, and in order to extract real value from them, you have to dig deep. Unlike traditional software, where prescriptive applications dictate what users need, AI operates differently—it’s a tool that requires more active engagement, iteration, and refinement to deliver the right output. Companies need to adopt a more “innovation” (rather than “invention” ) approach to using AI, as McKinsey point out in a recent article. Again it’s not “that” you use AI, it’s “how.”
The Pitfall of Prescriptive AI Apps
Most off-the-shelf GenAI solutions promise ease of use and instant value, but they fundamentally misunderstand the complexity of modern business functions—especially B2B sales. The reality is, most developers of these AI tools have little to no experience in B2B sales.
Here’s the problem: sales teams are expected to differentiate themselves in increasingly sophisticated ways. They don’t just need answers—they need nuanced insights enabling them to shape conversations, drive preference, and influence decision-makers. AI models that assume a one-size-fits-all approach will fail to deliver this, leading to frustration and rejection by sales teams who, after a poor first impression, may dismiss AI entirely.
Listen to the pod:
The Trade-Off: Quality vs. Speed in AI
One of the biggest challenges in developing AI-powered business tools is balancing response quality vs. response speed.
When using a private GPT for sales enablement, the amount of external data retrieved at query time impacts both response quality and speed. Expanding the indexed knowledge base generally leads to richer, more accurate answers but may slow down response times due to the retrieval process.
This means users often need to engage in a back-and-forth with AI to refine their results. While this is natural for those willing to invest time, it conflicts with the expectation set by off-the-shelf AI solutions: “It should just work.”
Most vendors prioritize speed over substance, leading to shallow, generic outputs. But in the details and nuances is where B2B sellers create differentiation—the very thing that drives deals forward. AI tools and the engagement models of those companies that fail to support this level of depth will be abandoned.
The CRM Fallacy: Inside-Out Thinking
Large software vendors like Salesforce are making similar mistakes. Their approach to AI is driven by an inside-out perspective—building AI as an extension of CRM rather than as an independent sales intelligence tool.
This results in AI that is:
Data-centric, not seller-centric
Process-driven, not insight-driven
Built for efficiency, not effectiveness
The assumption that AI should fit neatly into existing workflows ignores the reality of how top-performing sales teams operate. They don’t need an AI that simply summarizes CRM data—they need one that helps them think, strategize, and sell more effectively.
Now, this is not to say there isn’t some value in this approach. Prospects will need to decide whether value in the “platform” approach outweighs value in the standalone app. It’s ERP (or CRM) Vs Point Solution all over again.
AI Success Will Require Deep Client Collaboration
For the foreseeable future, the most successful AI solutions won’t be off-the-shelf products—they’ll be tailored, fine-tuned solutions customized with customers. Companies that want to harness AI effectively must invest time in:
Training AI models on their unique business context
Iterating on prompts and use cases to improve outcomes
Teaching teams how to interrogate AI for better responses
This means the subscription-based AI model—where companies expect an AI app to work flawlessly out of the box—is fundamentally flawed. The vendors who engage deeply with customers to refine their AI models will win.
The Funding Dilemma: Chasing the Wrong AI Model
The push for AI startups to scale fast is further complicating the landscape. VCs and private equity firms are chasing “unicorns”—betting on the first mover rather than the best mover.
But history tells us that being first doesn’t guarantee success:
Google wasn’t the first search engine.
Facebook wasn’t the first social network.
Apple didn’t invent the smartphone.
Microsoft wasn’t the first operating system.
Zoom wasn’t the first video conferencing tool.
The real winners in tech are the fast followers—companies that refine the business model, improve user experience, leverage ecosystems, or pivot at the right moment.
Many of today’s AI startups are overpromising and underdelivering in their rush to market. The result? A wave of AI fatigue as users lose faith in AI’s ability to drive real value.
The Future of AI in Sales & Business
For AI to truly transform sales, marketing, and other business functions, vendors need to shift their approach:
Stop promising magic. AI isn’t plug-and-play; it requires iteration and refinement.
Prioritize effectiveness over efficiency. Speed matters, but insight is what closes deals.
Engage deeply with users. The best AI solutions will be those built with the people who actually use them.
AI has the potential to revolutionize business, but only if it’s deployed intelligently. Companies that chase hype and shortcuts will fade. The ones that embrace depth, nuance, and real user needs will thrive.
Which side will you be on?
