AI governance gap: Between principles and practices
Governance is action. It takes the form of a wide array of planning, approval, access control, monitoring, remediation, auditing, and numerous other tasks that must be undertaken throughout the development, deployment, and maintenance lifecycle of every AI project. If these actions are not undertaken, there is no governance regardless of whether there are AI governance committees, principles, and policies in place or the number of risk audits that have been undertaken of the organization’s AI portfolio.
Key reasons why AI governance needs automation
Governance is disconnected
Risks are only assessed at the end of a project leading to rework, workarounds, delays or projects that aren’t put into production.
Evidence isn’t collected
Information on what data was used, how it was processed, what methods were used, and what libraries were called, is not captured in the moment and must be recreated after the fact for reviewers.
Results are not reproducible
Artifacts from the project, datasets, versions of code, versions of models, specific environments, and outputs are not automatically captured or available to reviewers — delaying their ability to replicate and validate solutions.
Policies are complex and unclear
Applicable policies are uncertain and change as the scope and risks of projects evolve.
Processes are unenforceable
Systems don’t exist to prevent or grant secure access to risky data, environments, infrastructure, or third-party services when risks materialize or are mitigated.
Fragmented ecosystems
The AI lifecycle spans a growing range of tools, technologies, environments (clouds, on-premises), types of infrastructure, third-party services (e.g. hosted LLMs).
Everything is done manually
All governance activities, ranging from documentation to coordinating approvals are done manually.
AI governance that is actionable and dependable
Domino Governance gives customers the best of both worlds — applying governance across all AI to mitigate risks instantly, without stalling innovation — so governance is actionable and automated. Now everyone has the time and ability to adopt best practices so governance becomes a catalyst to deliver more value from AI, faster.
Domino Governance helps enterprises:
- Accelerate impact: Streamline the model development and deployment processes
- Improve innovation: More resources and efforts dedicated to innovation
- Automate compliance: Seamless compliance documentation and reporting
- Mitigate risks: Continuously monitor and validate models in production
- Adapt swiftly: To new regulatory requirements and become a future-ready enterprise
Here is what effective AI governance in practice, at scale, looks like.
Check out the Domino Governance Maturity Assessment
Industries trust Domino to govern mission-critical use cases
Mission-critical work is also the riskiest. That is why Domino Governance protects mission-critical workloads to ensure trustworthy, safe, ethical and responsible AI by default. See why the most highly regulated industries trust Domino to govern their most urgent, mission-critical use cases.
Raj Mukerjee
Data Analytics and AI Director, Direct Line Group
“As a head of data science and AI, you've got to have a governance process that works across the business, because there are different types of AI. We’re incentivizing projects not to shy away from governance — and designing some of the platform work in Domino, where it’s within the development process by design.”
Ivan Black
Director of Engineering, Delivery Services, FINRA
“One of the big concerns, at least for us, again, FINRA being in a financial services space and being a regulator in a financial services space, privacy and data security are paramount. That's like the absolute one thing we can never mess up that has got to be perfect at all times.”
Chief Data Scientist
Allstate
“We can now go back to any project at any point in time and see what decision was made, and recreate that model if needed, which is huge in explaining to both management and regulators how we built a particular model — including what data went into it and what variables it identified.”
Francesco Maggina
Head of Data Science, ConTe.it - Admiral Group
“We had an extensive list of requirements for a data science platform covering DevOps, ModelOps, platform administration, privacy, security, training, support, and documentation. We found that Domino was the best platform due to its security, ability to go live with models quickly, and comprehensive features.”