3 Lessons from 2026 That Should Shape Your AI Strategy

In 2026, data and AI strategy sessions with clients started to sound different. Alongside questions about what AI could accomplish, discussions turned toward whether organizations could trust its answers, how to control costs, and whether AI could move beyond basic Q&A interactions.

These conversations revealed three core lessons that shape how organizations should plan their AI investments for 2027.

Lesson 1: Context is King: AI is Only As Useful As the Information It Can Access

AI models improved dramatically in 2026. However, even the best model can’t answer questions about your business unless it can reach the right information and understand its underlying meaning.

That context comes in two forms: structured data and unstructured data. AI needs both, but each presents its own challenges.

Structured Data: Consistent Definitions Matter

Structured data includes the metrics, definitions, and underlying tables behind your business reports.

We still see our clients struggling with the same problem. If “revenue” means three different things in your organization, which one should the AI assistant use?

That’s why semantic models and layers now sit at the center of AI readiness. A semantic model organizes business definitions, metrics, and relationships, while a semantic layer makes that context consistently available to AI and BI tools. Platforms like Databricks, Snowflake, and Microsoft keep adding features to help AI understand business data, but the technology can’t decide what revenue means or which metrics the business trusts. That’s still up to you.

What we’re seeing with our clients:

Unstructured Data: Making Organizational Knowledge Accessible

Documents are still where a lot of a company’s real knowledge sits: contracts, policies, technical standards, manuals, and correspondence.

Making that content useful to AI takes the same discipline as any data engineering work: ingestion, transformation, quality controls, security, and ongoing maintenance.

It also raises its own questions. Which version of a policy is current? Who’s allowed to see which contracts? How do you keep AI from quoting a document that was replaced two years ago?

What we’re seeing with our clients:

Lesson 2: Manage AI Costs: Get More Value From Every AI Dollar

As AI expanded into everyday business operations in 2026, the bills started adding up. Organizations began asking: Are we getting enough value from what we’re spending?

That question has put a greater focus on AI cost optimization, often termed “AI valuemaxxing”. The idea is to match models to specific tasks, keep an eye on consumption, and design applications that don’t waste effort. We’ve explored this challenge in more detail in our article on controlling AI spend.

Three important lessons emerged about how to get more value from AI investments:

• Match the model to the task: Not every task requires the most powerful model. Reserve frontier models for complex reasoning and use smaller, less expensive models for routine work like classifying documents, extracting information, or sorting support tickets. Using a frontier model for everything is like hiring a senior data architect to file paperwork.
• Automate model selection and reduce waste: Use an AI harness or gateway to automatically route requests to the appropriate models and manage how applications use tools, memory, and context. This reduces reliance on individual users or developers to choose the right model while avoiding unnecessary tool calls and processing.
• Monitor AI consumption and cost: Track which models are being used, by whom, for what purposes, and at what cost. This makes it easier to identify expensive or inefficient workloads and find opportunities to optimize spending.

The goal isn’t just to use cheaper models or reduce AI usage. It’s to make sure organizations are using the right resources for the right work, and getting enough value to justify what they’re spending.

What we’re seeing with our clients:

• Improving AI Cost Visibility: A large professional services firm with extensive AI adoption across its data platform asked us to conduct an independent review of its environment focused on cost visibility. The audit identified which teams and agents drove spend, giving the organization the information it needed to establish budget guardrails and identify opportunities to optimize workflows.

Lesson 3: Think Beyond Chatbots and Agents

For most people, AI still means a chat prompt: ask a question, get an answer. That’s useful, but it leaves a lot on the table. What if AI didn’t wait for someone to ask the question?

For organizations with established BI environments, it’s worth reconsidering which questions really need another dashboard or report. Standardized reporting will always have a role, but some analytical needs may be better served by AI that flags what changed and why it matters.

Proactive AI Workflow examples:

• Automated Portfolio Monitoring: A PE-backed organization growing through acquisitions told us its lean corporate team needed faster ways to spot problems across a growing portfolio of brands. Instead of waiting for month-end reports, leadership could get a weekly alert flagging which branches are falling behind on revenue per location, where customer cancellations are climbing, and which newly acquired brands still aren’t reporting clean data.

• Autonomous Lead Prioritization: We’ve built this kind of process for our own marketing team. Each week, an AI-enabled workflow reviews new marketing-qualified leads, pulling together their activity in HubSpot, previous interactions with Analytics8, open sales opportunities, and third-party intent data. It then prioritizes leads for follow-up and gives each seller a short summary explaining the recommendation, along with a relevant article or case study to share. It’s also designed to avoid making things up: recommendations are based on confirmed activity, and missing information is flagged rather than assumed. It runs automatically on a set schedule so nobody has to ask for it.

Turning 2026 Lessons into 2027 Priorities

If 2026 taught us anything, it’s that getting value from AI takes more than adopting new tools.

Organizations need trusted data and business context, a more intentional approach to managing costs, and a broader view of what AI can do beyond chatbots and conversational interfaces.

These lessons provide a useful starting point for 2027 planning. The question now is how organizations can put them into practice.

In our next article, we’ll explore three priorities for building a more effective AI strategy in 2027.