Generative ai has already transformed the world and it is just getting started. It has been rapidly adopted across multiple industries, from retail to healthcare and banking, and offers multiple capabilities ranging from information retrieval, expert help, and new content creation. With growing interest in GenAI from most boardrooms, CTO/CIOs are now faced with a big question: Should you build your GenAI app in-house or purchase a pre-built solution?
This article provides a framework to help product managers, business leaders, and technology leaders make this decision. Keep in mind that many of these fundamental arguments apply to all build versus buy decisions, but we present some nuances that are unique to the current landscape at GenAI.
The decision to build versus buy (vs. modify, which I consider under the purchase umbrella) depends on multiple factors. What makes the decision even more difficult is that the ai landscape is evolving rapidly, with new models and products being launched every week. What today may be a gap in market offerings could see a new product in the coming weeks.
The key factors that impact this decision are:
- Market availability (now and in the short term) and necessary commercialization speed of companies
- Strategic importance of the application to the business
- ROI for the business
- Risk and compliance factors
- Ability to maintain and evolve.
- Complexities of integration
In essence, my personal advice is to reframe the question of building versus buying as Why do you need to build? There are hundreds of incredible organizations investing billions into GenAI app development, so unless you are one of them, you should try to understand what is on the market that doesn't suit your needs.
- Unique business requirements: If your needs are unique, such that the applications available in the market do not fit your needs, and you believe that there will be no such applications in the market in the short to medium term horizon. Given the accelerated nature of development in GenAI, I have personally seen cases where organizations began building a feature, only to see it available as a commodity on the market within a few months. An example of this is the assessments, which saw many launches from many key players in 2024, including <a target="_blank" class="af qf" href="https://aws.amazon.com/bedrock/evaluations/” rel=”noopener ugc nofollow” target=”_blank”>AWS and <a target="_blank" class="af qf" href="https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/evaluations?tabs=question-eval-input” rel=”noopener ugc nofollow” target=”_blank”>Azure.
- Competitive advantage: If the application is of strategic importance to your business and is essential for you to maintain your intellectual property and your differentiation in the market. These would be very unique circumstances overall and would need to have strong leadership alignment for this. A well-known example here is that of LLMs. Most organizations do not need to create their own LLM models; They can use what is available on the market with well-designed prompts or adjust them to their own context. Bloomberg called for building his own modelwhich was a strategic move to enable the use of its proprietary data with finance-specific lexicon while consolidating its position as a leader in financial innovation.
- Long-term profitability: While initial development costs are higher, in-house solutions can be more cost-effective in the long run if used on a large scale. A common mistake here is not including long-term maintenance overheads in the cost when building the business case. It should also be noted that although many GenAI applications may be expensive now, the<a target="_blank" class="af qf" href="https://opusresearch.net/2024/07/29/costs-of-generative-ai-continue-to-drop-unlocking-new-possibilities/” rel=”noopener ugc nofollow” target=”_blank”> costs are falling rapidly As we speak, what may seem like an expensive purchase now may end up being cheap in a few months.
- Data privacy and security: Sensitive industries such as <a target="_blank" class="af qf" href="https://www.cio.com/article/3480467/how-to-build-a-safe-path-to-ai-in-healthcare.html” rel=”noopener ugc nofollow” target=”_blank”>health care and finance often has to comply with strict regulations and concerns about data privacy. In-house solutions provide additional control over data management and compliance.
If you ultimately decide to build in-house, some key challenges will arise:
- You will need a trained team of ai experts, a lot of time, and a significant upfront investment.
- Maintenance and updates, including compliance with the changing regulatory landscape, become your responsibility.
- Even with the right team of experts, you may not be able to keep up with the speed of innovation that currently exists in GenAI.
Pre-built GenAI solutions, available as APIs or SaaS platforms, offer fast deployment and lower upfront costs. Here's when buying might be the best option:
- Speed to market– If you're looking to deploy quickly, even if an existing solution may not 100% fit your needs right now. With new developments and releases, new functions can meet more needs.
- Predictable costs: Subscription-based pricing models provide absolute clarity in terms of expenses, avoiding cost overruns. On top of that, we have seen frequent price drops in GenAI and we expect that to happen in the short term. A recent example was that of <a target="_blank" class="af qf" href="https://aws.amazon.com/about-aws/whats-new/2024/12/amazon-bedrock-guardrails-reduces-pricing-85-percent/” rel=”noopener ugc nofollow” target=”_blank”>amazon Bedrock reduces prices by 85%.
- Focus on core priorities: Purchasing allows your team to focus on specific business tasks rather than the complexities of creating ai. This is especially true for solutions that are available as commodities and do not offer any competitive advantage.
If you ultimately decide to buy, some key challenges arise:
- You may be limited by how much you can customize. There may be features you need that remain in the vendor's backlog for longer than you'd like.
- Possible concerns about vendor lock-in and data privacy.
The build versus buy decision for each GenAI application should take into account the overall enterprise GenAI strategy to build versus buy. The decision cannot be made in isolation, as a critical mass of applications is needed to justify having a team to develop them. However, the following questions should be asked to help guide the response for individual applications:
- Will this application allow a clear competitive advantage?
- What solutions already exist on the market?
- What is your timeline for implementation?
- What capabilities are needed to build and maintain the GenAI application? Do you have that experience, either internally or with partners?
- What are your data privacy and compliance requirements?
- How is the business case different between building and buying?
The decision about whether to build or buy GenAI applications is not one-size-fits-all. The core of the decision-making strategy is not much different from any other build versus buy decision, but GenAI applications have the added complexities of a rapidly changing landscape, a high pace of innovation, and high but cost-reducing costs. relatively low. new technology.
While building offers control and customization at generally higher costs, purchasing provides speed and simplicity. Sometimes you may not have something available that fits your needs, but that can change in a few months. By carefully evaluating your organization's needs, urgency, resources, and objectives, you can make a decision that drives long-term success and value.
