Global automotive supplier TI Automotive operates at the heart of a rapidly transforming industry. The company is a market-leading Tier 1 supplier with advanced expertise in fluid management and lightweighting, supplying safety- and performance-critical products – from brake, fuel, and thermal management solutions to washer systems and interior & exterior components.

In this environment, data and AI are becoming increasingly central to engineering, operations, and decision-making across the automotive value chain.

Against this backdrop, in Part 1 of this three-part interview series, Apurva Wadodkar, Senior Director and Head of Data and AI at TI Automotive, explained what it takes to build an AI practice from the ground up, emphasizing the importance of education, prioritization, and disciplined execution.

In Part 2, Wadodkar speaks with Merav Yuravlivker, Chief Learning Officer at Data Society, about two foundational areas that determine whether AI initiatives succeed at scale: AI governance and decision intelligence.

AI governance is a shared responsibility

As organizations accelerate their adoption of AI, governance has become a central concern. According to Wadodkar, the first misconception many companies have is assuming governance belongs to a single team.

Instead, she argues that governance should be distributed across both a central oversight group and the teams building AI models.

“AI governance is not just one person’s or one team’s job,” says Wadodkar.

Her approach divides responsibility into two layers:

• A central governing body that evaluates use cases and establishes guardrails
• AI development pods dedicated for responsible model creation and validation

This structure allows organizations to maintain oversight without slowing innovation.

The 4 pillars of central AI governance

At the center of Wadodkar’s governance framework is a cross-functional council that evaluates AI use cases entering the organization’s development pipeline.

This central group operates across four key pillars:

• Security – ensuring models and data pipelines are protected
• Privacy – safeguarding sensitive and personal information
• Architecture – aligning development with enterprise platforms and standards
• Legal – evaluating regulatory or contractual risks

Wadodkar explains that this group reviews proposed AI initiatives and identifies potential risks early in the lifecycle. In many cases, governance can also be codified into clear rules to reduce friction.

For example:

• Prohibiting personally identifiable information (PII) from being used in model training
• Establishing standardized approval workflows for exceptions
• Requiring teams to document use cases through structured intake forms

“You can codify certain things,” Wadodkar notes. “Absolutely no PII data to be used for creating a model. If there is an exception, come to this meeting where people will hear you out.”

By embedding governance into processes rather than ad-hoc reviews, organizations can scale AI oversight without creating bottlenecks.

Why every organization needs an AI project repository

One of the most practical governance mechanisms Wadodkar recommends is a central repository of AI initiatives. It documents every custom AI project underway within the company.

This visibility becomes essential when customers or regulators ask how their data is being used. “Your customer can ask you anytime, ‘Hey, what are you doing with my data?’ And you need an answer.”

Without a clear inventory of AI systems and their data sources, organizations risk losing control over their own AI landscape. The repository also enables teams to track approvals across the governance pillars and ensure compliance before models move into production.

Architecture governance prevents reinventing the wheel

Beyond compliance and security, governance also plays an important operational role: preventing duplicated effort across teams.

The architecture pillar within Wadodkar’s governance structure helps enforce standards across AI development.

“We are not putting data in some awkward, different environment,” she explains. “We have a standard tool for this, let’s use it.”

Architecture oversight also ensures that teams collaborate rather than rebuilding solutions independently: “Certain teams might come up with a solution they’re about to build. But another team has already built it. So collaborate.”

This kind of architectural coordination allows organizations to:

• Reuse existing components
• Share utilities across teams
• Standardize development platforms
• Avoid fragmented tool ecosystems

“If the architecture team can go a little bit further, they would publish utilities to be reused and repurposed by all these teams.”

In large enterprises, these efficiencies can significantly accelerate AI adoption.

Governance at the Pod level

While central oversight ensures alignment and compliance, the teams building AI models retain responsibility for the quality and integrity of their systems.

Wadodkar refers to these teams as development pods, which work closely with business partners to identify and implement AI use cases.

This decentralized ownership is essential because the central team cannot manage every model across the enterprise.

Within these pods, developers must ensure several key responsibilities are met:

• Bias evaluation in training data
• Model explainability
• Accuracy and performance metrics
• Use case validation with business stakeholders

“Explainability is critical,” Wadodkar says. “There won’t be any adoption of your model if people don’t believe what you’re building.”

These responsibilities ensure governance is embedded directly into the model development lifecycle, rather than added afterward.

Decision intelligence: Turning data into better decisions

While governance ensures responsible AI deployment, Wadodkar believes the true mission of a data organization is improving decision-making across the business. That philosophy led her to develop a Decision Intelligence Framework, designed to align data initiatives directly with the decisions that shape company performance.

“Sometimes we get engrossed in becoming executors and less of partners, and that’s never good,” she says.

For Wadodkar, the role of a data leader extends beyond delivering dashboards or analytics pipelines. Instead, the goal is to enable better decisions at every level of the enterprise.

The foundation of the framework is deceptively simple: build a structured inventory of the most important decisions made in the organization.

“As a leader of your data organization, you make sure you always have a list of the most critical decisions being made in the company,” Wadodkar explains.

This requires collaboration with business leaders across departments.

“What are the top three finance decisions that are being made?” Wadodkar asks. “Profitability analysis, for example. Are the products we are producing and the customers we are selling to profitable? That’s a huge decision for the company. It changes the trajectory of your company. How is that not on your list?”

Building a decision intelligence loop

Once these decisions are identified, the next step is to map how data can improve them.

The framework captures key attributes of each decision:

• Decision description
• Frequency of the decision (annual, quarterly, bimonthly)
• Time currently required to make it
• Desired outcomes or metrics

With that foundation in place, data teams can then embed data products directly into the decision process.

“Now we are partnering,” Wadodkar says. “We are plugging in data products in each one of these decisions and creating a feedback loop.”

The process works like this:

• A decision is made
• Outcomes are measured
• Data products quantify results
• Insights improve future decisions

“We are tweaking that decision and making sure it’s better and better and better,” she explains.

One of the biggest benefits of decision intelligence is that it reframes how data teams measure success. Instead of producing reports or datasets, teams focus on improving decision outcomes.

“Now suddenly the data products are not just some tables,” Wadodkar says. “They are aiding that outcome.”

This shift helps data teams prioritize the initiatives that matter most.

“Build the right products,” she advises. “Don’t go for all the million things you could do.”

CDO Magazine appreciates Apurva Wadodkar for sharing her insights with our global community.