As your organization navigates the landscape of AI solutions and implementation, making the right choice between building in-house solutions or purchasing existing ones has become critical. The following framework is built from a panel conversation I held during the 2024 Support Experience Conference in San Jose. Three industry leaders joined me to share their expertise on the issue, bringing incredible experience not only on the operations side in post-sale CX, but in customer success, support digital channels, and the office of the CIO:

• Alvina Antar – CIO & Advisor, currently on the board of Couchbase and also an advisor for Signal and OluKai focused on the security and AI space. She was formerly CIO at both Okta and Zuora.
• Ashok Gunasekaran – SVP, Customer Success, Acceldata. Formerly of NetApp and Informatica, Ashok was instrumental at Informatica to building their own in-house customer sentiment solution that they used for many years.
• Vimal Vasudevan – Head of Digital Parcel USA, Koerber Supply Chain. Vimal was previously at Siemens Logistics for many years in product and technology roles.

You can watch the full discussion here on demand.

Step 1: Evaluate Your Business Needs

Begin by assessing whether the AI capability represents a core competitive advantage for your organization. As Alvina Anter, former CIO at Okta and current board member at Couchbase, explains: “The reality is you purchase technologies that are best in class, that are truly differentiated and that are critical to your business, and you invest in building those that differentiate your business that are part of your IP.” In other words, weigh the opportunity cost of building a solution and whether it fits within your business’s core competencies.

To evaluate business needs effectively:

☐ List your organization’s core competencies

☐ Identify areas where AI could provide competitive advantage

☐ Assess whether the AI solution needs to be uniquely tailored to your business

☐ Consider if existing solutions could meet your needs

Step 2: Analyze the Use-Case Complexity

Examine the specific use cases for AI in your organization. Ashok Gunasekaran, SVP of Success at Acceldata, emphasizes this point: “It all starts with the use case – what we are trying to change in the business model.”

When analyzing use cases:

☐ Document the specific business processes requiring AI

☐ Assess the uniqueness of your requirements

☐ Evaluate the complexity of implementation

☐ Consider industry-specific needs

☐ Determine data requirements and availability

Step 3: Calculate Your Internal Resource Requirements

Use Gunasekaran’s analogy when considering resource allocation: “Ideally I may want a custom home, but you have to spend a lot of capital on it, one, and the time that you get to construct is going to be very long. Then you are going to have to engage with an architect or a designer and a builder and a contractor and the responsibility falls on you to make sure that project goes per plan.
Yes, the end result may come, but it’ll be very long time. Versus you go buy it from a track builder, so you get a templatized version that they have done before.”

For accurate resource assessment:

• Calculate total cost of ownership for both options – Including not only the initial development or purchase costs but also ongoing maintenance, updates, and potential scaling requirements. For build projects, this includes developer salaries, infrastructure costs, training data acquisition, and the necessary computing resources.
• Estimate implementation timelines – Custom solutions typically require months or even years of development, testing, and refinement. In contrast, pre-built solutions can often be implemented within weeks or months, though customization and integration may extend this timeline.
• Assess internal expertise and availability – Building in-house requires a team with deep AI expertise, including data scientists, machine learning engineers, and domain experts. Even when buying a solution, organizations need staff who can effectively integrate and maintain the system. Training requirements should be carefully considered – both for the technical team and end-users who will work with the AI system.
• Consider training requirements – Resource availability must be evaluated not just for the implementation phase but for the entire lifecycle of the AI solution. This includes considering team bandwidth, potential hiring needs, and the impact on other projects. Factor in the opportunity cost of allocating resources to AI development versus other strategic initiatives.
• Factor in maintenance and upgrade costs – Custom solutions require ongoing development to stay current with advancing AI technologies, security requirements, and changing business needs. Purchased solutions typically include updates, organizations must evaluate whether these updates will align with their evolving needs and whether they’ll have sufficient control over the upgrade process.

Finally, consider the hidden costs and risks associated with both approaches. For build projects, this includes potential project delays, technical debt, and the risk of key personnel departures. For buy solutions, considerations include vendor lock-in, integration challenges, and potential limitations in customization capabilities.

Step 4: Apply the Three-Element Framework

Speed to market is crucial. As Gunasekaran notes, “At this time, everybody wants to show value faster. Nobody wants to wait for years to show an initial use case in the promise of AI. So I think there is a natural inclination in the current maturity of AI that people are trying to buy first.”

Follow Vimal Vasudevan’s approach from Koerber Supply Chain and structure your analysis around:

1. Assess the Job to Be Done	☐ Define the specific tasks AI will perform ☐ Identify the process improvements needed ☐ List the expected outcomes
2. Create the Problem Statement	☐ Clearly articulate the challenges ☐ Define the scope and limitations ☐ Identify your stakeholders
3. Build the Value Proposition	☐ Calculate the expected ROI ☐ Define your success metrics ☐ Identify the competitive advantages of each option

Step 5: Evaluate Each Solution

When considering existing solutions, use this six-step framework for making a complete evaluation.

Research Available Vendors	☐ Assess vendor track record ☐ Consider geographical presence and support coverage ☐ Check customer retention rates ☐ Evaluate support levels and SLAs ☐ Assess documentation quality ☐ Review training resources
Compare Feature Sets	☐ Create a detailed feature comparison matrix ☐ Identify must-have vs. nice-to-have features ☐ Evaluate customization options ☐ Compare AI model accuracy and performance ☐ Assess integration capabilities with existing systems ☐ Review scalability options
Customer Testimonials	☐ Contact reference customers ☐ Read case studies ☐ Check online reviews ☐ Evaluate industry-specific success stories ☐ Assess customer satisfaction metrics
Pricing Models	☐ Compare licensing structures ☐ Calculate total cost of ownership ☐ Assess upgrade costs ☐ Consider volume discounts ☐ Evaluate hidden costs

Step 6: Consider Data and Security Requirements

Use the framework below for a complete list of considerations in the data and security aspects of any AI solution:

Data Privacy Regulations	☐ Review GDPR compliance requirements ☐ Assess CCPA and other regional regulations ☐ Evaluate industry-specific requirements ☐ Check data residency requirements ☐ Review data handling procedures ☐ Consider cross-border data transfer implications
Security Requirements	☐ Evaluate encryption standards ☐ Review authentication methods ☐ Assess access control mechanisms ☐ Check security certifications (SOC 2, ISO 27001, etc.) ☐ Review incident response procedures ☐ Evaluate backup and recovery capabilities ☐ Assess vulnerability management ☐ Review penetration testing reports
Integration Needs	☐ Evaluate API security requirements ☐ Assess single sign-on capabilities ☐ Review data transformation needs ☐ Check real-time processing requirements ☐ Consider legacy system integration ☐ Assess middleware security
Scalability Requirements	☐ Evaluate performance under load ☐ Check resource allocation flexibility ☐ Evaluate cost implications of scaling ☐ Review geographic distribution options ☐ Assess multi-tenant capabilities
Compliance Considerations	☐ Review your industry-specific regulations ☐ Assess audit trail requirements ☐ Evaluate reporting capabilities ☐ Check certification requirements ☐ Assess monitoring capabilities

Step 7: Make the Choice

Based on your analysis, choose the approach that best fits your organization:

Consider building when:

• The solution represents core IP
• You have unique requirements
• Internal expertise is available for the long-term
• Long-term control is essential
• Custom integration is crucial

Buy when:

• Speed to market is crucial
• Resources are limited
• Standard solutions meet your needs
• Vendor expertise adds value
• Regular updates are important

Remember that this decision isn’t always binary. Some organizations opt for a hybrid approach, combining purchased solutions with custom elements. The key is to align the decision with your organization’s strategic goals while considering practical constraints of time, resources, and expertise.

By following these steps and carefully considering each aspect, you’ll be better equipped to make an informed decision about whether to build or buy your AI solutions in 2024 and beyond. Remember that the landscape is constantly evolving, so regularly reassess your strategy to ensure it continues to meet your organization’s needs.

Finally, click below to watch the full discussion from the 2024 Support Experience Conference and here more on the subject from our panelists.