A Framework for Tech Investment: Build Process vs Build Application vs Buy vs Use AI

Every week, I speak with executives trying to map out their technology and AI strategies for 2027. A common question I hear is, "Should we use AI for this?"

While Generative AI is incredibly powerful, it isn’t a silver bullet for every operational bottleneck. Sometimes you need an LLM; sometimes you need a predictive dashboard; and sometimes, you just need to buy an existing software license.

To help our clients avoid costly missteps, we use a simple framework to determine whether to use an LLM (like Google Gemini), build an analytics process, build a custom app, or buy off-the-shelf software. Here is how they compare across cost, timeframe, internal capabilities, and team alignment.

1. Buy an Application (SaaS / Off-the-Shelf)

When to use: The problem is a standard business process (e.g., CRM, HR, payroll) that does not provide a unique competitive advantage.

• Cost: Low upfront capital expenditure; predictable recurring operational costs (licensing).

• Timeframe: Fast. Weeks to implement and train.

• Internal Capabilities: Low technical maturity required.

• Functional vs. IT Work: Clear separation. IT handles vendor security, single sign-on, and access management. The functional business team owns the day-to-day workflow and administration.

2. Leverage LLMs (e.g., Google Gemini, ChatGPT)

When to use: You need to process unstructured data, generate content, summarize documents, or build basic workflows for internal applications.

• Cost: Very low barrier to entry for API usage or enterprise chat licenses.

• Timeframe: Immediate to extremely fast (days to weeks for basic RAG or prompt engineering).

• Internal Capabilities: Requires basic AI literacy and prompt engineering, but not necessarily hard coding skills.

• Functional vs. IT Work: Blurred lines. With LLMs, functional teams can act as "citizen developers," creating their own automated workflows and prompts. IT's primary role shifts to establishing data governance, managing API costs, and ensuring enterprise data doesn't leak into public models.

3. Build an Analytics Process

When to use: You have the software in place, but you need to extract proprietary insights/reporting from your historical data—like predicting customer churn, optimizing inventory, or segmenting your market.

• Cost: Moderate. Investment goes toward data infrastructure and specialized talent.

• Timeframe: Medium (weeks to months) to clean data, build pipelines, and validate models.

• Internal Capabilities: Requires data engineering and data science expertise.

• Functional vs. IT Work: Highly collaborative. Functional teams must define the business rules and KPIs. The data/IT team executes the pipelines and builds the models to deliver the insights back to the business users.

4. Build a Custom Application

When to use: The workflow is highly unique to your business operations and creates a distinct, proprietary competitive advantage in the market.

• Cost: High capital expenditure for design, development, and ongoing maintenance.

• Timeframe: Slow. Months to years for a full software development lifecycle.

• Internal Capabilities: Requires a mature product management, software engineering, and DevOps culture.

• Functional vs. IT Work: Heavy IT/Engineering lift. Functional teams act as the "client" or Subject Matter Experts (SMEs), providing requirements and user testing, while the IT/development team completely owns the technical build and architecture.

📌 Key Takeaway

Before committing budget, ask yourself: Does this workflow uniquely differentiate our business?

If no, buy it. If you need to summarize or create text, use an LLM. If you need numerical insights to make better decisions, build an analytics/BI process. Only build custom applications when it is core to your competitive edge and the use case is extremely unique.