The first meeting with a potential customer often starts with a broad statement: "We need to use AI." Your job as a Pre-Sales Engineer is to move beyond that buzzword. The goal isn't to sell AI; it's to solve a specific, costly business problem with the right technology. This is technical discovery: a deep dive to uncover the high-value opportunities where Generative AI can make a real impact.

Think of yourself as a detective, not a salesperson. You're looking for clues that point to operational bottlenecks, manual workflows, and hidden inefficiencies. The key is to translate a business pain point into a technical blueprint. This process starts with a structured approach. Instead of a free-form conversation, you use a framework to guide your questions and ensure you gather all the critical information needed for a successful project.

A solid discovery framework prevents you from getting lost in details and keeps the focus on what matters. While many sales methodologies exist, a Sales Engineer's perspective adapts them to map business needs to technical realities. One popular framework in enterprise sales is MEDDPICC, but for our purposes, we can distill the core technical discovery elements into three stages: quantifying the pain, auditing the data, and defining success.

First, you must quantify the pain. A vague problem like "our reporting is too slow" isn't actionable. You need numbers. Ask targeted questions:

• How many hours does your team spend on this manual task each week?
• What is the error rate of the current process, and what does each error cost the business?
• How much potential revenue is lost because of this bottleneck?

Attaching a number to the problem (e.g., "We lose 200 person-hours per week copying data between systems") transforms it from a complaint into a business case. This is the foundation for calculating ROI later.

Next, audit the data. An AI model is only as good as the data it's trained on. You need to investigate the customer's data landscape. Key questions include:

• Where does the data needed for this task reside? Is it in a single database, or spread across multiple data silos like Salesforce, SharePoint, and custom applications?
• What is the format of the data? Is it structured (like in a SQL database) or unstructured (like PDFs, emails, or call transcripts)?
• Are there any privacy or compliance constraints we must consider, such as GDPR, HIPAA, or CCPA? Access to sensitive data is often a major hurdle.

With a clear understanding of the pain and the data, you can start defining what success looks like. This means establishing measurable Key Performance Indicators (KPIs) before any solution is built. A good KPI is specific, measurable, and tied to the original pain point. For example, instead of "improve efficiency," a better KPI is "reduce average RFP response time from 40 hours to 4 hours."

This discovery process helps you identify high-value use cases that are a perfect fit for Generative AI. These are often tasks that involve summarizing, generating, or extracting information from large amounts of unstructured text.

The final output of your technical discovery is a clear picture of two states: the current, manual process and the future, AI-powered process. You should be able to map out the existing workflow, highlighting the exact points of friction. Then, you can design a 'to-be' workflow that shows precisely where the Generative AI solution will intervene and what its impact will be. This mapping is the bridge from a business problem to a technical solution architecture.

This structured approach ensures that when you move to the solutioning and architecture phase, you're building a system that solves a real, quantified business problem, not just implementing technology for its own sake.