About This Episode
In episode 8 of Data & AI Heads, host Ian Allison speaks with Charles Boyle, a seasoned Chief Data and Analytics Officer with more than 20 years of experience building and scaling Data and AI Centers of Excellence across industries including insurance, aerospace, and industrial technology. Charles breaks down the four primary pillars of a successful COE: people, process, data, and technology, and explains why getting each one right, in the right sequence, is critical to delivering business value at scale.
Charles draws on his tenures at Octo Telematics and Honeywell Connected Enterprises to offer practical guidance on topics that data and AI leaders wrestle with daily: how to prioritize a backlog of 100 ideas with a team resourced for 10, when to insource versus outsource specific roles, how to evaluate and migrate tool stacks without derailing productivity, and how to structure a hub-and-spoke COE model that works across decentralized business units. He also shares candid lessons from a rushed BI tool migration that delivered only 80% of its promise, and a rigorous platform evaluation at Honeywell that was ultimately overridden by CTO-level integration concerns.
For data and AI executives tasked with proving the value of their function, Charles outlines how he connects COE roadmaps directly to corporate strategic objectives, uses product-level P&L analysis, and maintains regular upward communication to keep C-suite stakeholders bought in throughout the year. This episode is a practical field guide for any leader building or maturing a data and AI capability inside a complex organization.
Key Takeaways
- A Data and AI COE rests on four interdependent pillars: people, process, data, and technology, and gaps in any one of them will constrain the performance of the others.
- Demand management is one of the most overlooked gaps in a COE: without a formal intake and prioritization process, executive ad hoc requests will overwhelm the team and dilute impact.
- Data architects and data scientists should almost always be kept in-house because they carry the intellectual property and tribal knowledge that create competitive advantage, while commoditized roles like data engineering and BI development are safer candidates for outsourcing.
- When evaluating a new or inherited tool stack, spend the first 90 to 120 days on a structured assessment across people, process, data, and technology before committing to any changes, and always factor in budget constraints alongside capability gaps.
- Tech stack integration is a critical but frequently overlooked criterion when selecting analytics platforms: even an independent, rigorous vendor evaluation at Honeywell failed to include integration compatibility in the scoring rubric, which required significant rework after the decision was made.
- A hub-and-spoke COE model, where the enterprise center sets best practices and recommendations while execution remains decentralized in business units, can reduce the cost of scaling while still driving consistency across the organization.
- Connecting the COE roadmap explicitly to corporate revenue and cost-savings objectives, and communicating progress on a quarterly basis, is the most reliable way to maintain C-suite buy-in and demonstrate return on investment.
- A product-level P&L that tracks both the cost of running data and AI solutions and the revenue or savings they generate gives leaders a defensible, concrete way to calculate and communicate ROI to executive stakeholders.
What We Cover
Frequently Asked Questions
What are the core components of a Data and AI Center of Excellence?
According to Charles Boyle, a Data and AI COE is built on four primary pillars: people, process, data, and technology. People are the most impactful component; data is the foundational layer that makes analytic and AI solutions possible; process governs how work is prioritized and executed; and technology tools provide the efficiency needed to operate at scale. Gaps in any one of these pillars will limit the effectiveness of the entire function.
Which roles should a COE keep in-house versus outsource?
Charles Boyle recommends insourcing data architects and data scientists because these roles carry intellectual property and competitive advantage that organizations need to protect. Commoditized roles such as data engineering and BI development are safer to outsource, as they involve more execution-focused work without the same IP exposure. This model has been successfully deployed across multiple organizations in his career.
How do you prioritize a large backlog of data and AI requests with a small team?
Charles Boyle addresses this by establishing cross-functional intake teams and data science councils that evaluate incoming requests against available data, appropriate methodology, and business impact before any work begins. The goal is to identify which initiatives among a large backlog, he uses the example of 100 ideas with capacity for only 10, will drive the most measurable return. Regular upward communication to L1 leadership ensures alignment and buy-in on those prioritization decisions.
How should a new data leader approach evaluating an inherited technology stack?
Charles Boyle recommends spending the first 90 to 120 days conducting a structured SWOT analysis across people, process, data, and technology before making any changes. Budget availability matters as much as capability: in some cases overpaying for legacy tools creates an opportunity to reduce OpEx and reallocate savings. He cautions against rushing to judgment on tool migrations, citing a BI platform switch to Power BI that took two and a half years to pay back because the lift-and-shift proved far more complex than anticipated.
What is a hub-and-spoke COE model and when does it work well?
In a hub-and-spoke COE model, the central enterprise team sets best practices, tools, and techniques, while execution remains decentralized within individual business units or operating companies. Charles Boyle used this structure at Honeywell Connected Enterprises, where it lowered the cost of scaling by distributing execution costs to the business units while still providing enterprise-level guidance. It works particularly well in large, multi-unit organizations where mandating centralized execution would create resistance or slow adoption.
How do you measure and communicate the ROI of a Data and AI COE to executive leadership?
Charles Boyle ties COE initiatives directly to corporate strategic objectives around revenue growth and cost savings, then tracks impact at the product level using a product-level P&L. This includes calculating cloud infrastructure costs, data storage savings from policies like data retention rules, and top-line revenue generated from monetized data products. Regular quarterly roadmap reviews with C-suite stakeholders keep leadership informed and eliminate end-of-year surprises about what the COE was working on and why.
What is the most common mistake organizations make when selecting a new analytics platform?
Based on Charles Boyle's experience at Honeywell, one of the most common oversights is failing to include tech stack integration as an evaluation criterion, even when using an independent third-party auditor to assess platforms. His team evaluated tools including DataRobot, Rapid Miner, and DataIQ with rigorous case studies, yet the final decision was overridden by the CTO because none of the top-ranked platforms integrated cleanly into the existing tech stack. He recommends making integration compatibility a formal, weighted factor in any platform selection rubric.
About the Guest
Charles Boyle
Charles Boyle is a data and AI executive with more than 20 years of experience building and scaling high-performing Data and AI Centers of Excellence across multiple industries. His most recent roles include Chief Data and Analytics Officer at Octo Telematics and Chief Data Scientist at Honeywell Connected Enterprises, where he led the creation of a hub-and-spoke COE spanning aerospace, industrial, and connected technology business units. Charles specializes in the democratization and monetization of AI and machine learning, with deep expertise in COE organizational design, analytic demand management, insourcing and outsourcing strategy, tool stack evaluation, and aligning data and AI roadmaps to corporate revenue and cost objectives. Throughout his career he has managed globally distributed teams including data engineers and data scientists across Poland, India, Romania, and South America.
Connect on LinkedIn →More Episodes