LISTEN
Listen to this perspective
Written by Oleg Cohen. Companion narration uses the Brian voice.
A business can have capable people, useful AI tools and an important growth priority, and still struggle to connect them.
Employees become more productive. Developers build faster. Vendors introduce new features. Individual workflows improve. Yet a consequential question remains: What has the company become reliably able to do, and what difference does that make to its business?
That is the question shaping my work at Kainora.
I believe one of the most important obstacles to realizing AI’s potential lies in how businesses organize the journey from an important priority to dependable operation. That journey crosses business judgment, company knowledge, engineering, systems, authority, adoption and ongoing support. Its success depends on those elements working together. Too often, the connections between them receive less attention than the individual pieces.
The priority has an owner. The technology has an owner. The path between them is fragmented.
Consider a familiar situation. Leadership wants growth. A business team identifies an opportunity. Analysts document requirements. A technology team selects an approach. Developers or vendors build something. Operations inherits the responsibility for making it useful.
Each participant may perform their assigned role competently. But important questions can remain unresolved between the handoffs. What exactly must the business become able to do? Which decisions can the system make? Where is human judgment required? Who handles exceptions? What must operating teams change? Who accepts the capability, and who establishes whether it creates value?
When these questions emerge late, the company pays through rework, delay, additional coordination and dependence on a few people who keep everything connected. In The Missing Owner I described the questions that fall between business leaders and technology leaders. This is what that gap costs once the work is under way.
AI can accelerate individual activities within that arrangement. The larger opportunity is to improve the arrangement itself.
The distinction is visible in the research. In McKinsey’s 2026 survey, nearly nine in ten respondents reported regular organizational AI use in at least one business function, and 80% reported improved individual productivity. Meanwhile, 37% attributed at least some enterprise earnings impact to AI. These are different measures, but together they illustrate the distance between widespread use, personal benefit and enterprise financial results.
My interpretation is that closing that distance requires continuity: a connected way to define, engineer, introduce and sustain valuable operating capabilities.
Start by defining the ability the business needs
Imagine a manufacturer that wants to grow while protecting margins and keeping delivery commitments.
AI could help prepare quotations faster. That is useful. The broader capability is the ability to make profitable commitments the company can fulfill.
Establishing that ability requires bringing together customer requirements, material costs, production capacity, supplier constraints, commercial rules and the people authorized to approve exceptions. It also requires understanding what happens when information is incomplete, costs change, capacity disappears or a customer requests something unusual.
Now the undertaking has a meaningful business definition. The company wants its people to make sound commitments, consistently, under agreed conditions. That definition guides the architecture, the technology choices, the operating responsibilities and the evidence needed to judge success.
It also creates room for growth. A dependable quoting capability might allow the business to handle more opportunities, respond to more complex requests or serve customers it previously struggled to support.
Connect the work from discovery through operation
At Kainora, we organize this undertaking around five connected phases:
- Envision: Understand the business priority, identify the valuable ability and assess whether the investment makes sense.
- Design: Define the knowledge, decisions, responsibilities, constraints and acceptance conditions that make the capability coherent.
- Build and connect: Engineer its realization using reusable foundations and the company’s relevant systems.
- Validate and roll out: Test ordinary work, difficult cases and recovery; prepare people; establish accepted use.
- Operate and evolve: Observe results, review exceptions, improve what is useful and decide what should change next.
Evidence can reopen an earlier decision. Operating experience may reveal a misunderstood constraint, an unnecessary approval or a valuable new possibility.
This continuity also needs continuity of meaning. A “profitable order,” an “approved exception” or an “available resource” should retain a clear definition as it moves from a business discussion into software and everyday work. Our approach combines AI’s ability to interpret context with explicit business meanings, rules, relationships and constraints. This gives engineering and operation a shared foundation.
AI belongs throughout the undertaking
AI can help examine company knowledge, expose inconsistencies and explore alternatives during discovery. During design, it can help examine scenarios, trace dependencies and identify missing requirements. During engineering, it can assist implementation, integration and testing. During operation, it can support decisions, coordinate authorized work and identify patterns worth investigating.
The opportunity is to connect those contributions so that useful knowledge carries forward between phases. Human judgment, review and accountability remain explicit.
This also changes how I think about speed.
The meaningful measure is time to an accepted, useful operating capability.
That measure includes the effort required from the customer, the rework needed to resolve misunderstandings and the work required to establish dependable use.
Keep the evidence connected to the investment
In the manufacturer example, three questions matter.
First, can the team reliably use the capability, including its exceptions and approval boundaries?
Second, what changes in the work? Perhaps quotation turnaround improves, fewer commitments require senior intervention or costly corrections decline.
Third, what is the business consequence? Does the company win additional profitable work, protect margin or use capacity more effectively?
Each question requires evidence. In The Work After the Agent Finishes I argued that “done” is a claim. So is “improved,” and so is “profitable.”
Time released becomes an economic benefit through what the company does with that capacity. Revenue growth must be considered alongside delivery cost and contribution margin. The investment assessment must include implementation, internal participation, ongoing operation and support.
A credible ROI discussion connects attributable benefits to those costs over a defined period. It also acknowledges other changes that may have influenced the result.
Make the complete undertaking accessible
For a middle-market business with a lean leadership team, assembling and coordinating every required specialty can become a substantial undertaking of its own.
Kainora brings business understanding, capability engineering, the AI-native Lattice platform and agreed operating support into one coordinated system. The customer contributes its priorities, knowledge, authority and operating participation. Kainora coordinates the agreed design and delivery responsibilities. Together, we establish what acceptance means and how the business effect will be assessed.
The work can begin with one consequential priority. Establish the ability. Put it into actual use. Verify the difference. Use that experience to decide what comes next.
That is what I mean by moving from AI adoption to intelligent execution: a business becoming able to accomplish something valuable, reliably, with evidence that it matters.
What must your business become reliably able to do for its next priority to succeed?
Sources and further reading
- McKinsey, The State of AI in 2026, August 2026.
- The Missing Owner and The Work After the Agent Finishes, earlier essays in this series.