Casey Kindiger is a technology entrepreneur and CEO of Droma (formerly Grokstream).
When a customer reports a service disruption, different parts of the business often see different problems. Operations sees an infrastructure alert, finance sees a billing exception, and account management sees a renewal at risk. Each team may have AI tools to interpret its information, yet connecting those perspectives still requires people to exchange messages, reconcile records and find someone who remembers a similar incident. AI is accelerating individual tasks, but it is also exposing a fundamental weakness in how enterprises operate: Intelligence remains fragmented, even as the technology becomes more capable.
As organizations prepare for 2027, closing that gap should become a leadership priority. A recent McKinsey article, citing its latest state of AI survey, reports that 80% of respondents say AI has improved their individual productivity, while only 37% report a positive impact on earnings before interest and taxes (EBIT). That gap illustrates the challenge ahead: Translating local improvements into better business outcomes requires changes to how information, decisions and responsibility move across the enterprise.
Having spent much of my career building operations automation and AI systems, I see four priorities for measurable AI maturity in 2027:
1. Shift The Focus From AI Adoption To Enterprise Intelligence
The cognitive enterprise offers a useful framework for this transition. The concept predates generative AI; IBM described it in its 2019 annual report as an organization informed by data and AI and equipped to adapt. Generative AI now makes institutional knowledge more accessible, while agentic AI can pursue objectives through multiple steps and systems. What is different now is the possibility of a more explicit cognitive architecture: systems that connect perception, institutional memory, judgment, governed action and learning across the enterprise.
These capabilities bring the vision closer, but their usefulness depends on the information and authority available to them. An agent cannot reliably resolve a customer issue when account records disagree, procedures are outdated or responsibility for approving a remedy is unclear.
After years of building businesses around IT operations and AI-driven automation, I’ve learned that identifying a problem is rarely the hardest part. The greater challenge is connecting what the organization knows to the decisions it needs to make—and ensuring those decisions lead to meaningful outcomes.
The organization also needs to connect technical evidence with business context and carry that understanding across teams as they decide what to do. Otherwise, each department can become more efficient while the overall process remains slow, inconsistent and dependent on a few experienced employees.
2. Connect Knowledge, Judgment And Action
A modern cognitive enterprise connects six capabilities: sensing changing conditions, retrieving institutional knowledge, evaluating possible responses, governing authority, acting within defined boundaries and learning from verified outcomes. Together, they support a continuous process in which new information informs decisions and the results improve subsequent work. This requires more than access to documents. Institutional knowledge includes why a previous response succeeded, which assumptions proved wrong and when an exception was justified. Making that context available helps teams evaluate unfamiliar situations without reconstructing every decision from scratch.
In the service-disruption example, a connected workflow could associate technical symptoms with affected customers, retrieve previous remedies, assess service commitments and recommend a coordinated response. A reversible technical action might proceed automatically, while a consequential service change would require approval. Customer-facing teams could receive the same verified account of what happened and what comes next. The resulting value includes less repeated investigation, fewer contradictory recommendations and shorter delays between identifying a problem and resolving its business consequences. Recording the outcome also gives the organization a stronger basis for responding when conditions change again.
3. Build Accountability Into The Workflow
As AI systems gain the ability to act across systems, governance becomes part of the architecture that makes useful autonomy possible, not a control layer added afterward. Each workflow needs an accountable owner, clear limits on automated action and a record of the evidence supporting consequential decisions. Access to information should not automatically confer authority to change a system or make a customer commitment. Human reviewers need sufficient context to evaluate recommendations, along with practical ways to pause, redirect or reverse execution.
My view, informed by work in operational AI, is that broader responsibility should follow demonstrated performance on the decisions a system will actually handle. A system that identifies missing evidence and requests assistance can be more useful than one that produces a confident response to every situation. Learning requires comparable discipline: Teams should verify an action’s effects before treating it as a precedent. Otherwise, an apparent success can reinforce a flawed assumption or conceal costs transferred to another department.
4. Measure Business Outcomes, Not AI Activity
Leaders can begin with one workflow that crosses departmental boundaries and has a measurable consequence, such as service restoration, customer onboarding or order exception handling. Map where work stalls, where employees repeatedly search for information and where handoffs lose context. Establish baseline measures for completion time, rework, errors and customer impact, then identify the knowledge needed to improve those results. That foundation should include authoritative sources, consistent definitions and named owners responsible for keeping information current. Employees who manage exceptions should help shape the design because their experience often reveals gaps in formal procedures.
Before enabling automated action, define what AI may recommend, what it may execute and when it must escalate. Evaluate results across the complete workflow, including downstream workload and oversight costs, before extending the approach. Teams can then extend the same information sources, decision rules and safeguards to related workflows, allowing enterprise intelligence to develop incrementally. The more important measure is whether each resolved case improves the starting point for the next one. AI maturity should increase not only execution speed but learning velocity.
In 2027, AI maturity won’t be measured by the number of agents deployed or tasks automated. It will be measured by whether the enterprise becomes more capable because of the work it has already done—turning experience into better judgment, safer action and faster learning.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?