Subscribe

Social Media Links

Insights

 | 6 minute read

AI Improves Performance When It Is Connected to Data, Workflows, and Decisions

Never Miss an Insight

This article is part of Ankura’s Digital & AI Value Series, exploring how organizations convert technology investment into measurable business performance. New articles are delivered every other week.

Sign up to receive each article directly in your inbox.

Executive Summary

Many companies can point to AI activity. Fewer can point to the measurable results it produced. The difference is rarely the model itself. Value is realized when AI is connected to trusted data, redesigned workflows, clear governance, and measured outcomes, so leaders should decide where AI must change performance first and then build the connected capabilities, controls, and adoption needed to capture it.

How 6 Levels of AI Capability Turn Technology Investment Into Measurable Results

The Performance Question: Where Should AI Change Results?

The practical question for leaders is not which AI tool to adopt. It is where AI can improve revenue, gross margin, productivity, working capital, decision quality, or risk, and what evidence would show that it worked. Enterprise AI is progressing beyond stand-alone chatbots and general productivity tools. Companies are connecting AI to trusted data, business applications, workflows, and decision processes. As these connections deepen, AI advances from helping employees find information to generating insights, incorporating external signals, assisting work, executing multistep processes, and continuously optimizing operations.

The six levels describe increasingly advanced capabilities, but they are not a mandatory implementation sequence. Companies can develop them selectively based on business value, technology and data readiness, risk, and the requirements of each use case. This creates a business decision, not simply a technology decision.

6 Levels of Enterprise AI Capability

Why Many AI Efforts Fall Short of Enterprise Value

Pilot failure is usually an operating and execution problem, not simply a technology problem. A technically successful pilot is incomplete if it does not change how decisions are made, how work flows, or how performance is measured. Four gaps recur.

Fragmented data and technology. Data that is hard to trust and systems that do not connect confine AI to isolated use cases that cannot be reused.

Workflows that do not change.  AI that sits beside the process rather than inside it produces insight that no one acts on.

Unclear ownership and controls.  Without defined decision rights, accountability, and human oversight, deployment slows and confidence erodes.

Benefits that are not measured.  When value is not defined before deployment and tracked afterward, investment is hard to defend and hard to repeat.

AI Performs Only as Well as the Capabilities Beneath It

The Digital Capabilities Model organizes the enterprise capabilities required to create value. The AI progression describes how sophisticated AI becomes within those capabilities. The two frameworks are related but serve different purposes. Intelligence, automation, competitive insight, and autonomy are stronger when they share governed data, integration, controls, and operating standards, so each use case should deliver an outcome while strengthening reusable enterprise capability.

Digital CapabilityAI ProgressionHow AI Contributes
IT FoundationSupports every levelArchitecture, integration, security, governance, and model oversight
Enterprise Data PlatformSupports every level, especially Level 1AI-ready data, semantic context, traceability, and reusable services
Enterprise IntelligenceLevels 1 and 2Natural-language access, insights, prediction, and recommendations
Process AutomationLevels 4 and 5Copilots, intelligent workflows, and agents
Competitive IntelligenceLevel 3External market, competitor, customer, and supplier intelligence
Autonomous FunctionsLevel 6Closed-loop decisions, execution, measurement, and learning

Natural-language search gives employees fast, consistent access to trusted answers, which can improve decision speed and reduce manual reporting effort. Employees ask questions in ordinary language and receive answers from trusted enterprise data, documents, and knowledge sources. A governed semantic or ontology layer supplies the definitions, relationships, and context needed to interpret business questions consistently. The practical question for leaders is whether the underlying data is trusted enough for the business to act on.

In Practice: A multi-location organization used AI to standardize inconsistent operating classifications, then enabled leaders to search and compare performance without continuously maintaining manual mapping files.

Level 2: AI-Powered Business Intelligence

AI-powered business intelligence moves beyond retrieving information to explaining performance, identifying exceptions, forecasting outcomes, and recommending actions. Generative AI works alongside purpose-built statistical and machine-learning models for forecasting, anomaly detection, classification, and optimization. Value is realized when these insights reach the decisions that drive margin, productivity, and cash, which depends on governed KPIs that finance and operations both trust.

In Practice: A services company combined financial, workforce, GPS, and timekeeping information to reveal labor inefficiencies, understand account profitability, and identify opportunities to strengthen margins.

Level 3: External Intelligence

External intelligence expands the company’s field of vision beyond internal systems. AI can continuously collect and organize competitor prices, supplier costs, product availability, customer behavior, weather, construction activity, economic indicators, and other market signals, then evaluate them alongside enterprise economics. The advantage comes from connecting those signals to internal economics so that pricing, sourcing, and demand decisions reflect the market, not only the company’s own history.

In Practice: A multi-location business combined internal product economics with competitor pricing and demand data to identify pricing opportunities and improve promotional returns and gross margins.

Level 4: Automation With AI Copilots

AI copilots assist employees with analysis, preparation, and coordination that previously required significant manual effort. Embedded within existing tools and workflows, they gather information, analyze exceptions, draft summaries, recommend actions, and route work for review or approval. Value depends on adoption: The copilot must sit inside the workflow, employees must trust and use it, and management retains review and approval authority.

In Practice: A clinical copilot analyzed patient messages and available information, recommended follow-up questions, and suggested potential actions for professional review, supporting faster and more consistent decisions while keeping clinicians in control.

Level 5: Automation With Agentic AI

Agentic AI advances beyond assistance to the performance of defined work. Agents interpret information, determine next steps, coordinate activities across applications, execute permitted actions, monitor completion, and bring employees in for approvals or exceptions. Because agents act, governance becomes a requirement rather than a preference. Leaders must define what an agent may do, which actions need approval, and how each step is traced and measured.

In Practice: An AI agent contacted customers due for follow-up service, conducted routine communications, scheduled appointments directly in the operating system, and transferred exceptions to employees.

Level 6: Autonomous Operations

Autonomous operations integrate intelligence, decisions, execution, and learning into a closed loop. The system senses current conditions, selects an action within established authority, acts through connected applications, measures the operational and financial result, and uses feedback to improve future decisions. Autonomy is a governed extension of enterprise capability, not a stand-alone novelty, and it is appropriate only where data, integration, controls, and measurement can support action within defined guardrails.

In Practice: An insurer automated claims review and prioritization using rules and analytics. The pathway toward greater autonomy adds permitted execution, outcome measurement, and continuous learning to this operating loop.

Human Oversight Evolves With AI Maturity

As AI assumes more responsibility, the human role shifts from collecting information and performing routine work toward setting objectives, approving consequential actions, supervising exceptions, and improving performance. Governance should enable responsible scale rather than appear only as a constraint, and human judgment remains essential wherever risk, exceptions, accountability, or material decisions require oversight.

AI MaturityPrimary Human Role
Levels 1 and 2Evaluate information, interpret insights, and choose actions.
Levels 3 and 4Assess recommendations in context and review prepared work.
Level 5Define authority, monitor execution, and handle exceptions.
Level 6Establish objectives, guardrails, thresholds, and performance expectations.

A Value-Led Path to AI Maturity

The path is iterative, not linear. Each use case should deliver an outcome while strengthening reusable capability, making the next use case faster and less costly to deliver.

StepAction
1  Define ValueIdentify the financial and operational outcome to improve.
2  Design the CapabilitiesDetermine the decisions, information, workflow, and capabilities required.
3  Select the LevelApply the AI capability appropriate to the opportunity and risk.
4  Build and ValidatePrototype with employees, and test usefulness, controls, and economics.
5  Embed, Adopt, and GovernIntegrate the solution into systems, workflows, and decision rights, supported by training and change management.
6  Measure, Reuse, and ScaleTrack realized value, reuse data and platforms for the next use case, and increase automation where results support it.

Practical Implications for Leaders

The practical question for leaders is what to decide, fund, govern, measure, or change.

Prioritize by value. Select the performance levers AI should change first, and define where value will appear in operations or the financial model.

Invest in reusable foundations. Fund the data, integration, and security capabilities every later use case will share.

Assign ownership and decision rights. Name a business owner for each outcome, and decide who approves, monitors, and handles exceptions.

Match oversight to autonomy. Raise controls, traceability, and human review as AI authority increases.

Measure and reuse. Track adoption and realized value against the business case, then decide where to scale or stop.

For CFOs, this is a capital allocation and value-tracking question. For CIOs and data leaders, it is an architecture and operating model question. For risk leaders, it is one of accountability and defensibility.

How Ankura Can Help

Ankura connects performance improvement insight with digital execution. We combine business and industry expertise, data and AI capability, and financial and economic discipline to help companies identify where AI can change results, select the capability appropriate to the opportunity and risk, and build solutions that are embedded in workflows, governed, adopted, and measured. The goal is to move from stalled pilots to enterprise-scale capability, with each use case strengthening the foundation for the next.

© Copyright 2026. The views expressed herein are those of the author(s) and not necessarily the views of Ankura Consulting Group, LLC, its management, its subsidiaries, its affiliates, or its other professionals. Ankura is not a law firm and cannot provide legal advice. 

Let’s Connect

We solve problems by operating as one firm to deliver for our clients. Where others advise, we solve. Where others consult, we partner.

I’m interested in
I need help with