Most companies treat AI as a technology project, and that is exactly why so many initiatives fail. Sid Bharath, founder of Refound, argues that the real barrier is not the technology itself but the absence of change management and a clear operating model. Without those foundations, AI lands on broken processes and delivers nothing.
Bharath built Refound to help mid-market companies become genuinely AI native. His approach starts with a deep audit of current workflows, identifies where friction lives, and then rebuilds processes around AI rather than stacking tools on top of legacy habits. The result is measurable lift, not shiny-tool fatigue.
Key Insights
1. Here is what you need to know about why AI initiatives fail.
AI projects collapse most often because companies launch large initiatives without preparing their teams for a new way of working. Leaders announce the change but skip the change management that follows. Workers are suddenly expected to interact with an intelligent system that behaves nothing like traditional software or a human colleague, and no one shows them how. That gap between announcement and adoption kills momentum fast.
2. Here is what you need to know about unbundling broken processes.
Overlaying AI on a flawed workflow just accelerates the flaws. Bharath’s team uses an unbundling approach: they break a process into its component tasks, examine each one, then rebuild the workflow with AI handling the right pieces. This method forces companies to confront what is actually broken before automation enters the picture, turning the AI rollout into a genuine redesign rather than a costly patch.
3. Here is what you need to know about becoming the orchestrator.
Most people use AI backward. They let the model do the thinking and then execute its instructions themselves. The correct model flips that relationship: the human sets direction and the AI acts. Bharath runs a personal AI environment he calls Jarvis, connected to his email, calendar, projects, and client files. He directs; Jarvis executes. That shift from doer to orchestrator is the core mindset change AI native work requires.
4. Here is what you need to know about phased AI adoption.
Trying to transform everything at once is a recipe for failure. Bharath recommends launching in phases: deploy a focused initiative, monitor usage, collect feedback, iterate, then expand. This approach keeps teams from feeling overwhelmed, surfaces problems early when they are still cheap to fix, and builds internal confidence. Each successful phase creates momentum for the next, compounding adoption across the organization over time.
5. Here is what you need to know about data readiness.
AI that draws on a company’s own data is only as good as that data. Most organizations have years of accumulated, uncleaned records that they never prioritized because humans could sort through the mess themselves. AI cannot. Bharath recommends running data cleanup in parallel with the AI build rather than waiting for a perfect dataset first. Starting with what exists and improving as you go keeps momentum without letting data debt indefinitely block progress.
Pull Quotes
“What’s the point of just overlaying AI on top of a process that’s probably broken in the first place? You didn’t accelerate a broken process.”
— Sid Bharath
“You should be doing the thinking, and you should tell the AI what to do. The AI should go and do it.”
— Sid Bharath
“If AI takes over, like say, even 90% of what you’re doing right now, which you don’t want to do, that frees up your time to start doing things that you’re more interested in.”
— Sid Bharath
“It’s got to come from the C-suite. They have to be the ones who are exploring these technologies and trying it out themselves and exploring a new way of working, and only then would it trickle out to the rest of the organization.”
— Sid Bharath
AI Adoption and Workflow Transformation: Key Statistics from Refound
| Statistic | Detail |
|---|---|
| AI initiative failure causes | Lack of change management and team preparation, not technology failure |
| Typical admin burden for sales reps | Up to 5 hours per week spent on presentation work instead of sales calls |
| AI task automation potential | AI can take over up to 90% of repetitive, manual, or admin-heavy tasks |
| Agentic AI capability timeline | Models became significantly more capable at autonomous task execution roughly 4–5 months before the interview (early 2025) |
| Primary audit finding | Most departments show high volumes of make-work: copying data between apps, generating reports, and managing dashboards manually |
| Recommended adoption approach | Phased rollout — deploy, collect feedback, iterate — rather than one large upfront delivery |
Related Resources
Our Host
John is the Amazon bestselling author of Winning the Battle for Sales: Lessons on Closing Every Deal from the World’s Greatest Military Victories and Social Upheaval: How to Win at Social Selling. A globally acknowledged Sales & Marketing thought leader, speaker, and strategist, he has conducted over 1500 video interviews of thought leaders for Sales POP! online sales magazine & YouTube Channel and for audio podcast channels where Sales POP! is rated in the top 2% of most popular shows out of 3,320,580 podcasts globally, ranked by Listen Score. He is CSMO at Coevera. In his spare time, John is an avid Martial Artist.
Comments