According to Inc, an MIT report from August revealed that custom AI solutions are being “quietly rejected” by businesses, with only 20% piloting enterprise AI and just 5% having fully deployed systems. Impel, which develops AI merchandising for auto dealers, now generates around $200 million in annual recurring revenue working with GM, Hyundai, and Honda after incorporating AI in 2020. Forethought creates AI customer support agents that learn from existing documents, while Grammarly’s AI writing assistant serves 40 million daily users across 50,000 organizations. Grammarly helped Databricks save $1.4 million annually by cutting editing time by 50% and boosting resolved support cases by 10-15%.
Sponsored content — provided for informational and promotional purposes.
The gradual approach that actually works
Here’s the thing about AI adoption that most companies get wrong: you don’t need to boil the ocean. These successful implementations all share a common strategy – they start small and build confidence. Impel literally begins by only activating AI responses outside work hours, then gradually expands as employees get comfortable. Forethought calls it “starting at a crawl” before increasing sophistication. This is the exact opposite of the typical enterprise software rollout where everything gets thrown at users at once.
And you know what? It makes perfect sense. When you’re dealing with technology that could potentially replace human work, you’re fighting against deep-seated fears and skepticism. The salespeople at car dealerships weren’t exactly begging for AI – they needed to see it wouldn’t take their jobs but would instead eliminate the “soul-crushing, mind-numbing work” as Impel’s founder put it. That’s a much easier sell than “we’re implementing AI across the organization.”
Solving real problems, not chasing trends
What strikes me about these examples is how practical they are. These aren’t abstract AI experiments – they’re solving specific, measurable business problems. Grammarly fixing communication errors that nearly went out to customers. Impel helping salespeople identify which leads are actually worth pursuing. Forethought handling routine customer questions so human agents can focus on complex issues.
Basically, they’re using AI to do the stuff nobody wants to do anyway. And when you frame it that way, resistance melts away pretty quickly. The Databricks example is particularly telling – they had a small editorial team manually reviewing everything, which is exactly the kind of tedious work that AI excels at handling. No wonder employees told Grammarly “Please don’t take this tool away from me, I can’t live without it.”
The ROI is real and immediate
Look, businesses don’t adopt technology because it’s cool – they adopt it because it makes or saves money. And these companies are delivering concrete numbers that would make any CFO pay attention. $1.4 million in annual savings for Databricks? 50% reduction in editing time? 10-15% more support cases resolved? These aren’t vague promises about future efficiency gains – they’re hard numbers that show up on balance sheets.
Impel’s growth from AI skeptic to $200 million in recurring revenue tells you everything you need to know. When salespeople start revolting at the idea of having the AI taken away, you’ve crossed the chasm from nice-to-have to essential business tool. That’s when AI stops being a cost center and starts being what Grammarly calls a “loss aversion strategy” – something companies can’t afford to lose.
What this means for the rest of us
So if only 5% of companies have successfully deployed AI, what can we learn from these outliers? First, start with your existing data and workflows – don’t try to reinvent everything. Second, focus on eliminating the work people hate doing. And third, demonstrate value quickly with small pilots before going big.
The MIT research might show widespread rejection of AI solutions, but these companies prove that when you approach implementation thoughtfully – when you respect people’s fears and demonstrate clear value – the technology sells itself. Maybe the problem isn’t that AI isn’t ready for business. Maybe businesses haven’t been ready for AI. But that’s changing fast.
