Design Jinja2 Report Templates with Data Aggregation from Assessment Components — lab audio overview

Host: Welcome back. You're in AI Solution Delivery — a course about how consulting teams actually ship AI projects for clients, from that first discovery conversation all the way to a working system. This chapter is about the very first step: use case discovery and data readiness assessment. In plain terms, that's the process of sitting down with a client, figuring out which AI ideas are worth pursuing, and checking whether their data is actually in good enough shape to build on. Today's exercise teaches you how to turn all that assessment work into a clean, professional report. Why does that matter so much? Expert: Picture this. You're a senior engineer at a consulting firm, and your team just spent two weeks running discovery workshops with a large retail client. You interviewed fifteen stakeholders, scored eight potential AI use cases, audited their data systems, and ranked everything by feasibility and business value. Now it's Friday afternoon, and the client expects a polished report on Monday morning — one that their executives will actually read and act on. Without a repeatable way to produce that report, what happens? Someone on your team spends the entire weekend copy-pasting findings into a document, formatting tables by hand, chasing down numbers, and hoping nothing got lost. Worse, every engagement reinvents the wheel. A five-person delivery team can easily burn twenty hours per engagement on report formatting alone. That's pure overhead — time that should be spent with clients, not wrestling with Microsoft Word. The skill you're about to build is how mature consulting teams eliminate that overhead entirely. Host: So this is exercise one of three in this chapter, and it's your starting point — no prior context needed. The chapter introduction walked through the big picture of discovery and assessment. Now we get specific. What exactly will we build in this first exercise? Expert: You'll build two things that work together. The first is a set of structured data containers — think of them as labeled boxes that hold all the information you gathered during the assessment. One box holds the scoring results for each AI use case. Another holds stakeholder input — the quotes, concerns, and priorities you captured in interviews. A third holds the data readiness findings — what systems exist, what shape the data is in, where the gaps are. These boxes enforce rules. If a use case is supposed to have a feasibility score between one and ten, the box will refuse to accept a score of fifteen. For this, you'll use a Python library called Pydantic — it's an open-source tool that lets you define data shapes and validates them automatically. The second thing you'll build is a set of report templates. A template is just a document with blanks in it — like a form letter where the names and dates get filled in automatically. You'll use a templating tool called Jinja2, pronounced "jinja-two." It's the standard template engine in the Python world, used everywhere from web frameworks to automation tools. Your templates will generate markdown — that's a simple text format with hashtags for headings and dashes for bullet points, the same format used in README files on GitHub. The key idea here is separation. Your data lives in one place, in those validated boxes. Your formatting and layout live in a separate place, in the templates. When the client wants a different look next quarter, you change the template, not the assessment logic. When you onboard a new type of assessment, you add a new box without touching the report layout. Host: That separation sounds clean on paper. Before someone starts coding, what's the thing that actually trips people up here? Expert: The tricky part is resisting the urge to put logic inside your templates. Templates should be almost boring — they pull in a value, they loop through a list, they show a heading. That's it. The moment you start doing calculations inside a template — averaging scores, filtering stakeholders, deciding which use cases are "high priority" — you've mixed your business rules into your presentation layer, and six months from now nobody will know where a number came from. So the rule is: do all the thinking, all the aggregation, all the ranking in your Python code, before the template ever sees the data. The template just displays what you hand it. If you feel yourself writing an if-statement inside a template that's more than one line long, stop — that logic belongs upstream. Host: Great warning. So let's land the plane. What will you actually be able to do after this exercise, and what comes next? Expert: After this, you'll be able to take raw assessment findings — scores, interviews, data audits — and turn them into a professionally formatted markdown report section, automatically, every time, with the same structure. You'll have a reusable foundation your team can drop into any client engagement. That twenty hours of weekend formatting work I mentioned earlier? It becomes twenty minutes. This is exactly the kind of production-grade building block that elevates a team from doing one-off deliverables to running a repeatable, scalable consulting practice. Next, you'll extend this by building the pipeline that assembles a full report — merging data from multiple assessment components into one unified document, handling the cases where findings overlap or contradict each other. After that, in the third exercise, you'll wrap it all in a service your colleagues can call on demand. For now, focus on getting the data shapes right and keeping your templates simple. That foundation is what everything else in this chapter stands on. Thanks for listening, and enjoy the exercise.