Earlier this summer, MIT released a landmark study, The GenAI Divide: State of AI in Business 2025. It is the strongest empirical research we’ve seen on why some organizations succeed with AI while most do not. The study analyzed 300 publicly disclosed AI initiatives, interviewed leaders from 52 organizations, and surveyed 153 senior executives.
In this month’s newsletter, we highlight the report’s key findings and add color from Cerulean’s experience.
Finding 1: Hype outpaces tactical progress for most companies
From the report:
One mid-market manufacturing COO summarized the prevailing sentiment:
"The hype on LinkedIn says everything has changed, but in our operations, nothing fundamental has shifted. We're processing some contracts faster, but that's all that has changed."
This aligns with what we hear from customers:
Many are encouraging their teams to use ChatGPT or similar, and are finding productivity gains in preparing pitches, analyzing data, or writing copy.
However, few have found an AI use case that fundamentally rewires a business process in a way that moves their bottom line.
At a recent in-person leadership workshop, we asked: “What level of impact has AI had on your business so far?” Here’s what they said:
No one said “None”, but nearly everyone rated the impact as “Possibly some, but immeasurable”. To move beyond this level into tangible impact, organizations should target repetitive, menial workflows that bog down operations.
Our experience shows that if you could teach a new hire to do the task with a clear right/wrong outcome, then you can almost certainly automate it with a tailored AI tool. This creates both financial impact today and momentum for future initiatives.
Finding 2: Address specific problems with tailored solutions
From the report:
Our research reveals a steep drop-off between investigations of GenAI adoption tools and pilots and actual implementations, with significant variation between generic and custom solutions.
A consistent pattern emerged among organizations successfully crossing the GenAI Divide…these organizations:
Demanded deep customization aligned to internal processes and data
Benchmarked tools on operational outcomes, not model benchmarks
Partnered through early-stage failures, treating deployment as co-evolution
Sourced AI initiatives from frontline managers, not central labs”
Cerulean’s experience paraphrases two of these essentials for success:
Measure the current operating cost of that process to substantiate ROI before automating.
Engage early users in design: instead of shielding employees from AI, involve them in shaping the solution. We find that workers welcome relief from menial tasks and become the strongest advocates for adoption when the tool embeds their insights.
Beyond that, we would also add two key learnings:
Identify “boring” processes e.g. manual paperwork, fragmented data, or repetitive steps that teams dislike but can’t avoid. Unsurprisingly, they are more excited to explore the art of the possible and develop a workable solution for these tasks.
Design for reliability: solutions must carry out the process correctly at least 95% of the time. General-purpose chatbots can “kind of” do the work, but achieving 50–70% accuracy leads to excess time spent checking the AI’s work, negating any benefits.
Since most organizations are early in their AI journey, the lesson is clear: effective solutions to specific problems win. By limiting scope, organizations can achieve visible success in weeks, not months.
Finding 3: System must be capable of improving
From the report:
Our data reveals a clear pattern: the organizations and vendors succeeding are those aggressively solving for learning, memory, and workflow adaptation, while those failing are either building generic tools or trying to develop capabilities internally.
The primary factor keeping organizations ‹on the wrong side of the GenAI Divide is the learning gap, tools that don't learn, integrate poorly, or don’t match workflows. Users prefer ChatGPT for simple tasks, but abandon it for mission-critical work due to its lack of memory. What's missing is systems that adapt, remember, and evolve, capabilities that define the difference between the two sides of the divide.
Users across organizations are deciding whether to trust in the efficacy of AI tools. At the same time, the tools are proving whether they are actually useful and reliable.
What we’ve found is that if an automation is excellent but not yet perfect — say 95% accurate — it makes a big difference if the tool is visibly improving over time. Organizations are more willing to invest effort into training and refining a tool if there is a clear trajectory toward near-100% accuracy.
If the tool isn’t improving, or worse, is incapable of handling a specific workflow, employees quickly default back to doing the task themselves. It becomes easier to ignore the tool than to struggle with its shortcomings.
Our key takeaway: a system must not only be tailored to the process it is automating, it must also improve over time. This visible progress builds trust, encourages continued use, and increases the chances of long-term success.
Finding 4: Look deeply at the back office
From the report:
Organizations that cross the GenAI Divide discover that ROI is often highest in ignored functions like operations and finance. Real gains come from replacing BPOs and external agencies, not cutting internal staff. Front-office tools get attention, but back-office tools deliver savings. Despite 50% of AI budgets flowing to sales and marketing, some of the most dramatic cost savings we documented came from back-office automation. While front-office gains are visible and board-friendly, the back-office deployments often delivered faster payback periods and clearer cost reductions.
We would add a few nuances here.
If your business relies heavily on relationship building to generate commercial activity, this takeaway is even more relevant. AI is perfect for automating menial, repetitive tasks — but it cannot replace humans for “relating” work, as we discussed in “Humans vs AI – who will do what jobs?”.
On the other hand, if your business runs an outbound model — for example, reading construction reports to identify new opportunities — there is a clear opportunity to automate rote tasks like lead qualification, which frees sales staff for higher-value activities.
Overall though, our experience is consistent with MIT’s findings: we’ve seen ROI well in excess of 10X come from the back office. Operations such as data entry, information gathering and document reconciliation are ideal targets because they are repetitive, measurable, and tied to well-documented costs.
Example ROI: If a process takes 20% of each employee’s time, then across a 10-person team that equals 2 full-time employees’. That’s capacity you can immediately reinvest.
Importantly, back-office improvements also strengthen the front office. Salespeople depend on back-office functions to respond to customers, but bottlenecks slow the cycle. By automating these tasks, sales teams respond faster, improving customer experience, and increasing conversion.
Finding 5: Get external help
From the report:
In our sample, external partnerships with learning-capable, customized tools reached deployment ~67% of the time, compared to ~33% for internally built tools. While these figures reflect self-reported outcomes and may not account for all confounding variables, the magnitude of difference was consistent across interviewees.
We’d add the nuance that internally built tools can succeed — for example, we’ve seen strong implementations of contract review, COI management, database querying, and AI-assisted time entry. But we’ve also seen internal projects stall out after six months with nothing to show for the effort. The reality is that success is not guaranteed, whether tools are built internally or externally.
That said, the data suggests external partnerships are more likely to reach deployment, and we suspect three main reasons:
Single accountability: Internal teams juggle AI alongside day jobs; external partners own the outcome.
Clearer goals: External projects carry explicit costs, forcing leaders to define scope, criteria, and timelines up front.
Expertise where it’s needed: Every company is great at something, but for most, that “something” isn’t AI. External partners bring the talent and focus to build systems that actually work and improve over time.
When it comes to out-of-pocket cost, yes, that is real — but it’s only half of the equation. The other is the value of successful automation. If the benefits far outweigh the costs, the investment decision is straightforward. In our experience, tailored AI tools deliver exceptionally high returns and offer a compelling alternative to relying on humans handling menial tasks they don’t enjoy.
Conclusion
From the report:
For organizations currently trapped on the wrong side, the path forward is clear: stop investing in static tools that require constant prompting, start partnering with vendors who offer custom systems, and focus on workflow integration over flashy demos. The GenAI Divide is not permanent, but crossing it requires fundamentally different choices about technology, partnerships, and organizational design.
For any leader seriously considering how to integrate AI into their business, this report is essential reading. It distills empirical evidence from hundreds of deployments and surfaces the drivers of real success. Most importantly, it shows what leaders must do differently to avoid stalled pilots and move toward measurable impact.
At Cerulean, we see the same themes playing out every week: strong interest but limited progress, ROI showing up most clearly in back-office functions, and success hinging on tailored systems that improve over time. We’ll continue to share field insights alongside the latest research to help leaders realize tangible returns from AI investments.