What would compel 95% of businesses to invest in something with zero ROI to date? Written down, it sounds objectively ludicrous, but it’s happening everywhere. Organisations have drunk deeply from the corporate Kool-Aid on a global scale and, arm-in-arm, fallen into the AI Hype Trap. The visceral gap between a priori visions and the reality for most businesses begs the question: how is anyone meant to win the futuristic AI race if we don’t have the knowledge, systems, and capacity to get there?
Hype Against Hard Facts
In October 2024, Salesforce seemed poised to reinvent the wheel with its latest innovation: Agentforce. Positioned as the kindling to spark the next era of enterprise AI: a suite of autonomous agents that could take action across service, sales, marketing, and more. Their vision was that every employee would soon have a digital co-pilot, the ultimate work buddy that would make everything easier.
“We are entering a new era of digital workers – autonomous AI agents that can take action on their own and augment the work of humans.”
Marc Benioff, CEO of Salesforce, TIME Magazine
However, a year on, the cracks are starting to show. What was supposedly a canon event in the timeline of Generative AI has produced decidedly lacklustre results. In November 2025, Business Insider reported that, internally, of roughly 12,500 Agentforce customers, fewer than half were paying for the product, and less than 2% of Salesforce’s total customers were having more than 50 Agentforce conversations per week.
Employees described a rollout marked by steep complexity and confusion, while another Salesforce administrator admitted that “normal businesses with normal admins don’t have the expertise to set this up – it’s too new for anyone to be an expert.” Together, they paint a picture of a product racing ahead of its ecosystem: a sophisticated tool released into environments without the infrastructure, skills, or clarity needed to use it effectively.
What we see is an underlying assumption of universal readiness that didn’t and still doesn’t reflect reality.
Growing Pains
Agentforce’s obstacles illustrate what MIT researchers call the GenAI Divide, that is the gulf between adoption, integration, and effective deployment of artificial intelligence across the business landscape. These challenges are industry-agnostic, reinforcing the truth that what we need right now isn’t more technology, but a sharp attitude adjustment.
Moreover, collated metrics mirror what we’re seeing on the ground. MIT’s report states that around 60 % of organisations have experimented with generative AI, but only 5 % have reached production scale, and roughly 95 % have yet to record any measurable ROI. In what other field would this be admissible?
The underlying causes are familiar – and exactly what agile consultancies like Chesamel are built to address:
- Skill Scarcity: Few teams have the expertise to implement or maintain AI systems.
- Culture Drag: Legacy processes stifle experimentation.
- Siloed Infrastructure: Data and workflows aren’t connected.
- Leadership Pressures: Transformation launches are for appearance’s sake, not with the operational health of the organisation in mind.
“Most enterprise AI tools don’t learn, don’t adapt, and don’t fit existing workflows.”
MIT Report 2025
The Agile Alternative
If Agentforce had been rolled out through an agile, embedded approach, the story could have looked very different. A nimble transformation partner would have:
- Started small: one workflow, one region, one clear outcome.
- Embedded multidisciplinary teams inside client operations to observe friction first-hand and iterate accordingly.
- Built live feedback loops between users, engineers, and leadership to track adoption data in real time.
- Measured ROI by outcomes.
- Scaled by evidence, expanding based on proven functionality.
Furthermore, MIT’s research shows that organisations working with external partners on AI are roughly twice as likely to reach production as those relying solely on internal teams.
That’s the difference between working in theory and building something that makes sense in practice. Organisations would benefit from the introduction of AI not just as a tokenistic technology implementation to keep up with the competition, but as a tailored capability that genuinely transforms how people work.
“Agility isn’t about moving faster, it’s about learning faster and embedding change where people actually benefit.”
– Chesamel
The Takeaway
Agentforce isn’t the only example of organisations running faster than they can walk. The same storyline replicates and multiplies across sectors – launches to fanfare followed by thin adoption and frantic efforts to retrofit the basics.
It isn’t about one singular company or product – it’s about mindset. Big business still equates investment with impact, confusing the purchase of a technology with the mechanics of transformation.
The organisations that will thrive aren’t the ones making the loudest announcements, but the ones learning fastest from smaller, smarter bets.
If you’re navigating your own transformation challenges, or wondering where to start, let’s talk.
Disagree with our take? Join the conversation on LinkedIn or share this article to your network and see what your peers think.
