AI readiness in the public sector

AI readiness in the public sector

Responsible and effective use of AI in the public sector isn’t just about buying the “right” product. Before making a purchase, organizations must understand how different types of AI – from traditional machine-learning models for data analysis to generative AI for content creation – all have different use cases, technical requirements, and oversight and governance needs.

AI readiness means organizations must be equipped to:

• Conduct targeted risk assessments and impact evaluations for specific use cases
• Define clear, purpose-specific use cases and enable teams to experiment, iterate, and learn
• Maintain appropriate human oversight, auditability, and accountability where it matters most

Procurement's role is centered around ensuring the right solutions are available when teams need them. Meanwhile, a lot of the work to build readiness happens where AI is actually deployed and used. For example, leadership sets the strategic vision for AI adoption and creates frameworks that allow teams to move more quickly on low-risk applications while ensuring appropriate oversight for higher-impact use cases. Your agency or department project team needs to really understand the use case or problem they are trying to solve, and think through change management to support adoption. Your technology team should handle the technical integration and ongoing support.

The procurement team supports this readiness by creating pathways for innovation while supporting responsible and good implementation. This means instituting agile procurement practices, setting up processes to streamline access to low-risk AI solutions, and building vendor relationships to better understand the market. For individual procurements, the procurement function also drives a buying strategy that takes into account the organizational mission, program needs, and implementation of responsible best procurement practices.

Since many public sector organizations are still building their AI capabilities, this section outlines how procurement can accelerate AI-readiness rather than slow it down.

Many public sector organizations are still building these AI capacities. In this section, we provide an overview of how to get your organization AI-ready, and the role of procurement in supporting AI-readiness.

What an AI-ready organization looks like

There are several enabling conditions that can help you and your organization get the most from your AI technology. The more of these capacities you have in place, the better you will be able to procure and adopt AI technology. It’s still feasible to buy AI without all of these conditions in place – and many public sector organizations do! – but the solutions are less likely to be used or used to their full potential, leading to wasted resources or possible harms to public sector mission outcomes.

If your organization is not as AI-ready as you would like, we invite you to consider how you might support building these capacities. This might look like advocating for an institutional AI adoption strategy, requesting more agile procurement practices to enable experimentation, having conversations with AI industry leaders to learn more and support market research, or dedicating staff time to upskill on AI.

Institutional AI readiness

Institutional AI readiness is established at the leadership level, often by people that work across public sector agencies, such as a Department of Technology, Department of Innovation and Technology, or Chief Information/AI Officer, Digital Transformation Agency. While procurement officials don't set this strategic direction, they play a crucial role in bringing it to life by creating purchasing pathways and building vendor relationships.

Creating and using a robust AI adoption strategy can help public sector organizations align on overall goals for using this new technology and accelerate institutional AI-readiness. A good AI adoption strategy will balance opportunities and risks: it should articulate both the vision for leveraging AI to improve public services and operational efficiency, and establish governance and oversight mechanisms to mitigate risks at the point of use.

By creating frameworks that enable innovation while ensuring appropriate controls, an AI adoption strategy can help agencies move forward confidently with AI adoption while managing potential downsides. The strategy should address key risks, including technical, operational, ethical, legal, and reputational considerations, as well as set out clear requirements for data privacy, transparency and accountability, security, bias monitoring, and regular risk assessments, as well as consider integration with existing legacy technology.

The strategy should establish clear criteria for categorizing AI systems based on their risk level and intended use. As part of this, procurement should be enabled to streamline access to low-risk technologies while ensuring appropriate oversight mechanisms are in place during implementation. This approach empowers teams to innovate while ensuring governance happens where it's most effective.

Critically, an AI adoption strategy should also include provisions for safe experimentation environments where teams can trial AI solutions, learn from pilots, and build internal capabilities before full deployment. This means procurement processes should support both rapid access to low-risk AI solutions for experimentation, and structured pathways for scaling proven solutions.

Data management is another critical component of organizational AI-readiness. Data management is the practice of organizing, cleaning, and governing information so it can be trusted and used effectively. Data engineering is the behind-the-scenes process of making sure all of this data – especially from fragmented or legacy systems – is collected, standardized, and delivered in a usable format for AI. In procurement, when legacy systems lack a data management and engineering plan, AI systems may produce misleading analysis or miss critical risk indicators, introducing procurement and compliance risks.

Good data management matters for many reasons:

• Procurement risk: Legacy systems often store contract data, pricing details, or vendor performance records in inconsistent formats. Without careful data engineering, AI may misinterpret or ignore this data, leading to flawed risk assessments or compliance gaps.
• Data silos: If information is spread across multiple, unconnected legacy databases, AI cannot see the full picture – introducing blind spots in supplier evaluations or spend analyses.
• AI dependency on structure: Unlike humans, who can sometimes infer meaning from messy records, AI requires structured and standardized inputs. Data engineering creates that structure so AI models can function.
• Future scalability: Agencies that invest in strong data management and engineering now will be able to more easily modernize legacy systems and unlock advanced AI capabilities for procurement in the future.

Department or agency project team AI-readiness

The department or agency project team that is buying AI technology knows their needs best, and if and how AI can help them achieve better outcomes. The procurement team should be involved and consulted, and can support the project team through providing purchasing frameworks, conducting market research, and building relationships with vendors. To help departments and project teams become AI-ready, they can:

• Develop leadership expertise and structures. Supportive and knowledgeable leadership is critical for teams to meaningfully build their own AI readiness.
• Build a strong understanding of the problem you are trying to solve. This problem should drive the technology and purchasing decisions. AI might not be the right solution to meet your needs.
• Know AI solutions: Teams need to understand not just their problem space but also which type of AI solution – generative, agentic, or simpler AI technologies – best fits their needs and context.
• Know your data. Seek to understand your own data. Good quality data is important for good AI outputs, but may not be available for all teams at the beginning of the project. Consider your data classification requirements and ensure your AI solution will be ready to transmit and analyze the data you wish to use. Procurement can help ensure that your vendor contract addresses your specific data requirements.
• Understand your capacity to actively manage your vendor relationships as part of your project implementation. Proactive contract management and close vendor collaboration are important for projects, especially those with technology deployments. AI solutions require ongoing monitoring and performance assessment. By closely monitoring progress and maintaining open communication channels to address issues early, you can help ensure that your AI solution is working as intended. To ensure a successful collaboration, it’s important to check that your internal team has the capacity and responsibility to manage the vendor’s work following the contract award.

Procurement AI-readiness

Procurement directors and procurement officials can do a lot to build readiness to support the implementation of responsible and effective AI.

• Understand what AI technology you can purchase already. Depending on your jurisdiction, you may be able to acquire AI solutions through framework agreements, and/or preexisting centralized or decentralized contracts. For example, your project team might already have access to CoPilot if your organization uses Microsoft Office, or Gemini through Google Workspace, or Amazon Q if your organization uses Amazon Web Services. This existing access also provides valuable opportunities for teams to experiment and build AI literacy before making larger procurement decisions.
• Embrace agile and innovative procurement approaches. This might be one of the most important steps that you can support to ensure better deployment and outcomes from AI. Enable your project team to buy differently and use outcomes-based contracting, rather than rigid requirements, to promote creative solutions and maximize value. Starting small and opting for iteration when possible can help manage risks and achieve the best results for AI purchases, as well as open the door to engaging with smaller, innovative providers that don’t typically participate in public procurement. You can also create umbrella contracts to hire design and delivery teams, and structure them in sprints. These approaches also enable learning from pilot projects to inform larger AI investments and strategy. (Depending on your current policies and practices, we know this can be challenging. Check out our Hacks for better AI procurement section for ideas on how to be more agile and iterative.)
• Support cross-organizational collaboration. Double-check you have the right people in the room. With technology like AI, your key stakeholders should be involved early and often. Your technology team is especially important – they should get involved as soon as a project team decides they need to make a purchase. Security and legal experts, and someone who can represent the voice of whoever will use or be impacted by the technology, also need to be consulted along the way as you develop requirements and evaluate proposals and vendor claims. Usually, the project team is responsible for pulling everyone together, but encourage them to include procurement early on in strategy meetings rather than waiting to make a purchase decision, and let them know if someone is missing.
• Build AI procurement expertise within your team. Consider training procurement staff on AI basics and the dynamic sector landscape. This helps procurement professionals ask better questions and provide more strategic support to project teams.