Our AI Strategy: Use case Framework (part 1)
Just last week, I spoke with a CDO who told me, 'We can’t sit on the sidelines anymore. We need to figure out how to onboard the company to AI. It’s here, it’s happening, and we need to have a plan’. This sentiment isn’t new; over the past year, I’ve heard countless variations of the same concern. AI feels like both an opportunity and a risk, and companies need a structured way to take advantage of it. That’s why I developed the following 2-part AI strategy:
A use-case driven framework (the “what”)
A lifecycle-based governance model (the “how”).
The framework simplifies AI adoption by breaking it into three tiers of use cases—Enterprise, Vertical, and Innovation—and is operationalized by the governance model, serving as the playbook for delivering AI solutions.
Today we’ll share the framework (part 1) and next time we’ll cover the governance model.
The Use-Case Driven AI Framework
A 3-tiered AI Framework centered around company specific use-cases
I break down AI use cases into three tiers, each reflecting a different way to integrate AI within your organization.
Enterprise (Organization wide) AI
Think of this as introducing your workforce to generative AI systems, such as ChatGPT, Claude, or Gemini. This layer focuses on training teams, establishing company-specific best practices, setting up security protocols, and creating policies that ensure responsible, effective AI use across the organization. Years from now AI will fade into the background of our daily lives and won’t require as much explicit training (similar to Word processors or “the cloud”), but until then, this module empowers your workforce with knowledge, guardrails and confidence.
Investing in this layer will enable your workforce to be more productive and efficient. Here are some examples:
Email drafting and refinement: Folks can spend more time ensuring their emails cover all action items coming out of a meeting, and less time ensuring the right levels of concision and grammatical accuracy.
Document synthesis: Summarize knowledge base articles and learn the basics of new technologies really quickly.
Virtual ideation: Using GenAI to work through your thoughts, improve a presentation, or sharpen your ideas.
Chat-based knowledge retrieval: Use your organization-wide AI to search through an internal knowledge repository for useful information instead of pinging colleagues or spending time sifting through it yourself.
All of these, and more, will help your workforce be more productive and efficient with their time, shifting the distribution of what they do from low-skill (e.g. spell check) to high-skill (e.g. creating a thoughtful presentation).
Vertical (Domain-Specific) AI
Beyond generic productivity tools, business units need AI solutions tailored to their specific goals. Some examples:
Cursor or Github Copilot for Software Engineering teams
Grammarly or Jasper AI for Marketing teams
Salesforces Einstein or Seamless.ai for Sales teams
Zendesk AI or Sierra AI for Customer Support
Vertical AI takes into account data and workflows unique to those domains. Vendor assessments, model validation, and parallel proofs of concept (POCs) happen here. It’s where you ensure each team gets the right tool for the job.
Examples of the benefits of investing in your Vertical AI capability:
Social Listening & Sentiment Analysis: Using AI to monitor brand mentions across social media and identify trends, competitor moves, and customer sentiment in real time.
Software development: Using tools like Cursor or Co-pilot to make developers more productive, enabling them to focus on the hardest parts of their job, like designing performant systems or ensuring code runs as expected in production.
AI-Driven Customer Support Chatbots: A support team implements Zendesk AI or Sierra AI to handle routine inquiries (password resets, troubleshooting steps), escalating advanced cases to live agents as needed.
Innovation (R&D) AI
This is the realm of highly specialized applications. Dedicated AI engineers and data scientists build, deploy, and continuously monitor custom models for performance, bias, and drift. In this tier, your team will build bespoke AI-powered solutions, heavily informed by your datasets and specific context:
A computer vision model to detect anomalies in your assembly line, enabling your QC engineers to focus on systems analysis and analytics on your process
A company-specific lead scoring engine that outperforms the generalized model in your CRM, by better targeting leads likely to convert, improving revenue
Building and integrating data models into your BI tool, improving its understanding of your analytics data, making the chat interface yield much better results for stakeholders, improving how people leverage data across the organization
The Use Case Continuum
As indicated above, the applications that fall within each of the tiers of the pyramid vary in scope and complexity. That said, they share important traits: all AI applications are non-deterministic and highly sensitive to data inputs, prompting, and user interpretation. Small upstream changes—like slightly different training data or a revised user prompt—can cause major shifts downstream. That might mean a malformed email accidentally goes out to your entire company, or a customer-facing chatbot gives incorrect information, leading to negative online reviews.
All of this underscores the need for a robust, holistic governance model to pair with the use-case driven framework to help companies leverage and adopt AI within the organization.
In our next piece, we outline our governance model, which is embedded into the AI development lifecycle, and how we apply it to every tier of the framework.