Build a Meeting Assistant in 12 Minutes: One WhatsApp Message Handles All Your Pre-Meeting Prep

From a Single WhatsApp Message to Meeting Prep

Imagine this: you have your seventh meeting of the day coming up, and all the relevant context is scattered across calendar entries and a long email thread. Every time, you spend precious minutes hunting it down. One Reddit user shared his solution — compressing the entire process into a single message: "Hey Tara, get me up to speed on my meeting with Andreas."

He built a lightweight meeting preparation agent on the SCALAN AI platform. The assistant automatically locates the corresponding event in Outlook Calendar, retrieves the relevant email thread, and sends back a concise briefing via WhatsApp: what the meeting is about, where the last conversation left off, and which questions are still unresolved. The whole thing took 12 minutes to build.

Why This Use Case Deserves Attention

The real highlight isn't the feature itself — it's how little infrastructure it requires. The author deliberately chose a "small but practical" scenario, and that's exactly the kind of scenario that best demonstrates the maturity of today's AI workflow tools.

The build process breaks down into three steps:

• Connect data sources: Integrate Outlook email and calendar, giving the agent the ability to read context.
• Describe the routine in natural language: Write out in plain English "what to check, what to include in the briefing, and how to handle missing information." The instruction itself is readable text, not code.
• Enable WhatsApp and link your account: Push the output to an app you're already using every day.

No custom backend. No separate automation service to build and host. The author specifically emphasized that this "routine" is human-readable text that reliably produces the same structured output every time. If you want to change the briefing format or the preparation rules, just edit the text — and that's fundamentally why it can be done in 12 minutes.

The Significance of Natural Language Orchestration

In the past, implementing similar automation typically meant calling multiple APIs, writing glue code, and handling auth and hosting — a "simple requirement" could easily eat up hours or even days. In this case, the logic is defined as a readable instruction text, representing a new paradigm for workflow construction:

Declarative, not imperative. The user describes "what result they want" rather than "how to achieve it step by step." This lowers the barrier for non-technical people to participate in automation, and makes rule maintenance transparent — you can directly see the instructions the agent follows, rather than black-box logic buried in code.

Stable and reproducible. The author notes that output is "reliably consistent every time," which is critical for real-world use. Many LLM-based tools struggle to gain traction due to unstable outputs; constraining tasks within a clear routine framework is a pragmatic way to improve reliability.

Delivered inside existing habits. The information isn't pushed to yet another panel you have to open separately — it arrives in WhatsApp, an app already deeply embedded in daily life. Delivering useful context at the moment you need it, in the place you're already looking — this design philosophy often determines whether a tool gets used consistently, more so than any feature list.

---

The distinction between Declarative Programming and Imperative Programming comes from computer science. Imperative programming requires developers to describe "how to do it" step by step — for example, first call the calendar API, then retrieve emails with filter conditions, then concatenate strings to generate a summary. Declarative programming only requires describing "what you want," with the system handling the underlying execution details. SQL is a classic declarative example: you write "find records matching these conditions" rather than "iterate through each row and evaluate." Using natural language as an orchestration medium pushes this idea to its logical extreme — even formal syntax is eliminated, and instructions are just sentences written by ordinary people. The tradeoff is a high dependence on the underlying model's reasoning capabilities; the benefit is an extremely low barrier to participation and extremely high maintainability: changing business rules no longer requires modifying code or going through a deployment process — editing a piece of text is enough.

---

Real-World Applications and Open Questions

Pre-meeting prep is just a starting point. Similar patterns can extend to many lightweight scenarios: reviewing history before following up with a client, organizing leads before a sales call, summarizing context before escalating a support ticket. The common thread is that the data already exists somewhere — it just lacks a low-cost way to aggregate and deliver it at the right moment.

The author ends the post with two open questions worth considering for anyone thinking about adopting tools like this:

• What would it take for you to actually use this before every meeting?
• What additional context would you want it to pull (e.g., CRM records, meeting notes, shared documents)?

These questions point to the true value boundary of such agents: the more complete the context it pulls, and the more seamless the trigger, the more likely it crosses from "cool demo" to "can't live without it."

---

CRM (Customer Relationship Management) records are among the most commonly mentioned additional data sources for this type of meeting prep agent. CRM systems (such as Salesforce and HubSpot) store structured data including customers' purchase history, communication logs, contract status, and sales stage. For sales or customer success teams, being able to automatically pull up "what concerns this customer raised last time and when their current contract expires" before a follow-up call often has a more direct impact on conversation quality than an email summary. Integrating CRM with calendar and email as a three-way data source is the key leap that takes these agents from "personal productivity tool" to "team workflow infrastructure" — but it also introduces complexity around data permissions, multi-system OAuth authorization, and more. This is precisely the technical and organizational tension behind the original author's open question about "what else would it take."

---

A Pragmatic Observation

This case study's persuasive power comes precisely from its smallness. It doesn't claim to disrupt anything — it simply eliminates a small, daily, tedious friction point. For anyone evaluating AI workflow tools, rather than chasing grand visions of full automation, it's worth finding a clear entry point like "pre-meeting briefings" — one with well-defined boundaries, obvious value, and easy verification. Being able to get it running in 12 minutes and replicate it reliably is itself the best proof of a tool's usability.

(Note: The original post's screenshots show a German-language interface, displaying the actual briefing content, the underlying instruction text, and the WhatsApp connection screen. This article is based on the author's written description.)