How to Build Your AI Decision Partner That Fights Decision Fatigue (Part 2)

From concept to working system: the complete technical walkthrough with copy-paste workflows (Guest)

Thanks for reading! To access our community, full prompt library, coaching, and AI tools saving leaders 5-10 hours per week, check out our Premium Hub.

---

Building a house requires both vision and blueprints. In Part 1, we explored the vision: AI as a thinking partner that helps leaders structure decisions without replacing human judgment. Now it’s time for the blueprints.

Think of this like having a master craftsman walk you through building your own workshop. You’re not just getting the finished product; you’re learning the principles, seeing the connections, and understanding how each component serves the larger purpose. The result? A custom Decision Assistant that speaks your language and adapts to your leadership style.

In Part 2, we’ll build together:

• Voice and text input capabilities through Telegram integration

• AI transcription systems that turn spoken problems into structured text

• The AI decision engine with custom prompts and conversational flow

• Automated decision logging and tracking in Google Sheets

• Complete workflows and prompts you can copy-paste immediately

Jakub Slys brings his engineering expertise from building scalable systems at Pega, Nokia, and Sabre to show us exactly how to construct this AI thinking partner. Using accessible no-code tools, he’ll demonstrate how any leader can create their own decision support system. For more of Jakub’s automation insights, check out iam.slys.dev.

This is the most technical piece I’ve shared yet. We’ll walk through automation workflows, API integrations, and system configurations. But here’s why I want you to push through, even if you’re not technically inclined: understanding how these systems work will fundamentally change how you think about AI’s role in leadership. You don’t need to become a developer, but seeing the mechanics will help you lead more effectively in an AI-driven world.

Here is Jakub.

---

Meet the Decision Assistant Agent

Our goal: set up a system where a leader can speak or text their dilemma (via a messaging app) and get back a structured decision analysis, courtesy of AI, within minutes. We also wanted it to feel conversational and supportive, not just a form to fill out.

And here is a glimpse of the final product!

Here’s how we built it, step by step.

Using Telegram as the “front end” - voice or text input

We chose Telegram (a messaging app) as the interface for our Decision Assistant.

Why? Because it’s instant, simple, and supports text and voice messages. Sometimes a leader might prefer to just talk out loud about a problem rather than type a long question - especially if they’re on the go or it’s late and they’re too tired to write. With Telegram, they can do either.

Telegram Trigger: In n8n, we added a Telegram Trigger node that waits for new messages from our bot. This is connected to our Telegram bot’s API (set up via BotFather with an API token; it is vital to provide the right Webhook URL endpoint while configuring the bot). The moment you send a message to the bot - whether it’s text or a recorded voice note - this workflow kicks off.

Detecting Voice vs. Text: The next step is a Switch node we nicknamed “Voice or text?” that simply checks the incoming Telegram message: does it have a voice file or just text?

This switch has two outputs:

• If it’s a voice message, the workflow routes to a chain that will handle audio.

• If it’s text, we route directly to the AI agent (skipping audio processing).

Transcribing voice messages (so AI can understand you)

If the user spoke their dilemma as a voice, we need to turn that into text for our AI to process. Thankfully, we have OpenAI’s Whisper - a highly accurate speech-to-text model. We use n8n’s integration to OpenAI for this.

Get Voice File: First, a Telegram node uses the file_id from the message to download the voice recording from Telegram.

Transcribe with Whisper: Next, an OpenAI node (configured for Whisper’s speech-to-text) takes that audio and transcribes it to plain text. In a few seconds, we get the user’s question in written form. For example, a spoken message like “I’m torn between promoting Jane or hiring someone new for the lead role” would come out as text very close to what was said.

This way, whether the manager typed or spoke, we now have their decision question in text form ready to feed into the AI Decision Assistant Agent.

---

Brought to you by COZORA👇. Get up to 50% off with the coupon in the Premium Hub.

---

Guiding the AI with a custom prompt (“Decision Helper” persona)

Now comes the core of the system: the Decision Assistant Agent itself. In n8n, we used a special AI Agent node, which allows us to create a conversational agent with a given prompt, memory, and tools. We configured it like so:

Prompt (User message):

We crafted a prompt that essentially says: “You are “Decision Helper”, a conversational agent for leaders. You guide them through decisions in a simple, structured way… [cut]”. This prompt ensures the AI knows its role and the format we want. It’s not going in blind - we specifically instruct it to produce that nice breakdown (Pros, Cons, Risks, Next Steps). The agent will first use this prompt to guide the conversation. If the user’s question is incomplete or ambiguous, it might ask a follow-up via the chat.

Decision Helper Agent (Prompt)

System message: We provided the current date.

OpenAI Chat Model: Under the hood, our agent needs a large language model to do the heavy lifting of natural language processing. We added an OpenAI Chat Model node (in our case configured to use a GPT-4 model) and connected it to the agent. This is effectively the “brain” of the assistant, providing the AI responses based on our prompt and the conversation so far. The model is powerful enough to synthesize answers in a structured way - listing bullet points of pros/cons, etc., as we’ve asked.

Memory for Context: To make the interaction conversational, we included a Memory node (a “buffered window” memory) that keeps track of the last 50 messages in the session. This means the agent will remember what was said earlier in the chat. For instance, if the AI asked “Are there any specific constraints or goals you have?” and the user answered, the agent will remember that detail when formulating the final analysis. The memory is keyed per user (using the user’s Telegram ID), so each manager’s chat is kept separate and persistent. This makes the experience feel much more personalized, as the AI can reference things you mentioned previously (“Earlier you noted the budget is tight…”).

Google Sheets Tools. Using n8n’s Google Sheets node, our workflow appends a new row to a Google Sheets document every time the assistant produces a decision report. The row includes columns like: a unique ID, the question asked, the pros, cons, risks listed by the AI, and a field for outcome. This essentially builds a decision journal. Over time, you have a simple dataset of past tough calls, which can be incredibly useful for reflection or even training purposes.

With the model and memory in place, the Decision Assistant agent node orchestrates the conversation. From the user’s perspective, it works like this:

Leader (you): “I’m torn. Should I promote Jane (an internal candidate) to the vacant team lead position, or hire someone from outside?”
Decision Assistant: “Sure, I can help with that decision. First, can I ask: What’s your top priority in this choice - speed, cost, team morale, expertise... or something else?”
Leader: “Good question. The biggest factor is speed - we need a leader ASAP to get the project back on track. Budget is also a concern; we can’t spend a fortune hiring”.
Decision Assistant: “Understood. So, promoting internally might be faster and cheaper, while hiring externally might bring more experience. Let’s break it down. Here’s what I see:
- Restated Question: Deciding whether to promote Jane (internal) vs. hire an external candidate for the team lead role.
**- Pros of Promoting Jane: She’s already familiar with the team and projects, so virtually no onboarding time. Likely boosts team morale (shows growth opportunities). Lower hiring cost and quicker fill since it’s internal.
**- Cons of Promoting Jane: She doesn’t have prior leadership experience; there might be a learning curve. You might still need to fill her old position. Team might perceive favoritism if others wanted the role.
**- Pros of Hiring Externally: Bring in a fresh perspective and possibly specific leadership experience. No internal power shifts or backfills needed.
**- Cons of Hiring Externally: Costlier and slower - recruiting could take 1-2 months plus onboarding. Cultural fit is unknown; there’s a risk they might not gel with the team.
**- Risks & Considerations: If Jane is promoted without training, the project could suffer if she struggles (risk to timeline). If a new hire doesn’t work out, you lose time and money and might demoralize internal team members who wanted a promotion.
**- Next Steps (if Promoting Jane): Maybe set up a mentorship or training plan to support her transition. Communicate to the team why she was chosen to keep things transparent.
**- Next Steps (if Hiring Externally): Define the exact skill gaps you need filled (to justify the hire) and involve the team in interviews to ensure cultural fit. Consider a contract-to-hire trial if possible.
Hope this helps! Remember, there’s no perfect answer, but now you have a clearer view. You’ve got this!”

Notice how the assistant kept a conversational tone - it even asked a clarifying question upfront - and then delivered the analysis in a friendly but structured format. This is exactly what we aimed for: streamlining the framework process (pros, cons, etc.) and presenting it in a human way. The manager can now review this and make a more informed decision (and perhaps even sleep better, having offloaded those swirling thoughts).

Sending the AI’s answer back via Telegram

Once the AI (ChatGPT model) generates the structured response, our n8n workflow takes that output text and sends it right back to the user on Telegram.

Telegram Response. We use a Telegram node (configured with the bot and chat ID of the user) to send the answer. From the user’s perspective, the assistant is just chatting in the Telegram thread. The response typically arrives within a minute or two of asking their question (depending on whether there were follow-up questions). It feels like a fast, thoughtful chat with a helpful colleague.

This immediacy is key. The whole interaction stays in a medium the leader is comfortable with (a chat app), so there’s no complex UI to learn - just ask and get answers, even in voice form if you want. It’s almost like leaving yourself a voice memo and having it intelligently answered.

All together: The workflow in n8n

To summarize, here’s how the pieces connect in the n8n workflow (visualize nodes connected in this sequence):

1. Telegram Trigger - Listens for new messages from our bot.

2. Switch (Voice or Text?) - Checks message type.

   • If Voice: -> Telegram Download File -> OpenAI Whisper Transcription -> (transcribed text)

   • If Text: -> (just use the text as is)

3. Decision Assistant Agent - Receives the text (from either path), uses the system prompt, ChatGPT model, and memory to generate a structured response (and possibly interact with the user for clarification).

4. Telegram Send Message - Sends the agent’s response back to the user.

And of course, we have the OpenAI Chat Model, Memory and Tool configured and linked into the agent node internally (so the agent knows which LLM to use and where to store chat history).

Everything above was built without writing a single line of code - just using n8n’s visual interface to drag nodes and set their parameters. The “logic” (like the content of the AI’s advice) comes from the carefully designed prompt and the AI model’s capabilities, not from hard-coded rules.

Real scenarios: how the Decision Assistant helps

To put this into perspective, let’s walk through a couple more scenarios where a tool like this shines:

• Hiring vs. Training Dilemma: Suppose you’re a team lead and you must decide whether to hire a new developer to speed up a project or train your existing team to cover the skill gap. You tell the assistant your situation via voice on your commute home. It responds with a clear list of considerations: the pros of hiring (extra capacity, new skills) vs. pros of training (cheaper, growth for current team), the cons of each (hiring takes time and budget; training might delay the project and overburden current staff), risks (what if the hire quits? what if training isn’t sufficient?), and next steps (maybe hire a contractor as a middle ground, or start a quick training program and set a checkpoint to re-evaluate). You arrive home with a structured plan forming, instead of a mental fog.

• Strategic Direction Choice: Imagine you’re a product manager deciding between two big feature initiatives for next quarter. Both have potential, but you can only pick one to focus on. You message the assistant: “Should we focus on improving the existing user experience or build new features to attract a broader audience?” The AI might ask for your company’s current goals (e.g. retention vs. growth). Then it would outline: Option A (UX improvements) - pros (strengthen core user base, quicker wins), cons (may not expand market), Option B (new features) - pros (could tap new customers, PR buzz), cons (spreads team thin, current users might feel neglected). It would list risks like Option B could flop if new features don’t find product-market fit, and suggest next steps such as conducting a quick user survey or an A/B test. This structured thinking could be the basis of your proposal to the exec team, and you pulled it together in minutes with the AI’s help.

• Personal Career Decision: It’s not just for team decisions - maybe you’re a manager deciding whether to pursue an MBA. The assistant can help you weigh the investment vs. payoff, risks (financial, time away from family), and next steps (talk to alumni, try free courses first) in a neutral way, which might be easier than discussing with a colleague where emotions could cloud things.

In all these cases, the Decision Assistant Agent doesn’t make the decision for you - but it acts as a coach or sparring partner, ensuring you’ve thought through key angles. It’s available 24/7, doesn’t get annoyed if you ask it 10 variations of the same question, and it consistently applies the structured approach. And because it’s conversational, it can actually increase your confidence in the process (“I haven’t forgotten to consider anything major; I’ve done my due diligence with the help of my AI buddy”).

From overwhelmed to empowered: conclusion

Let’s revisit our manager burning the midnight oil. With a tool like the Decision Assistant, her 10:47 PM crisis could have looked very different. Instead of staring at a blank notepad and second-guessing herself, she could fire up a quick chat with the assistant, streamline her thoughts, and get a clearer picture of her options in a matter of minutes. By guiding her through a framework and offloading the tedious parts of analysis, the AI lets her focus on what matters - her insight into her team and her leadership values - when making the final call.

This journey - from leadership struggle to structured framework to AI-powered solution - highlights an exciting truth: technology can augment our humanity, not diminish it. As a leader, you’re not handing over your decisions to a robot. You’re simply using a power tool for your mind. Just as spreadsheets don’t make the financial decisions but help crunch the numbers, an AI thinking partner doesn’t decide, but it does arm you with organized information and perspectives. In the end, you are still in charge, hopefully with a bit more clarity and a bit less stress.

If you resonated with this scenario, consider giving something like the Decision Assistant a try. You don’t have to be a coder to build it - tools like n8n make it accessible with low-code building blocks. (And if you are tech-inclined, you can customize it endlessly - different prompts, different integrations, maybe logging to a Notion database or triggering follow-up reminders to revisit a decision.)

The takeaway for any new manager or curious leader is this: You don’t have to navigate the fog of tough decisions alone. Structured thinking helps, and now we have AI to help with that help. It’s like having a mentor or coach who’s always available, tirelessly turning your uncertain questions into clearer choices.

So next time you’re agonizing over a decision, remember that a friendly bot and a solid framework might be just a message away. Embrace the partnership of AI and automation - not to hand over your responsibilities, but to amplify your capabilities. Fewer sleepless nights, more confident days. After all, the goal isn’t to make perfect decisions (spoiler: no such thing); it’s to make well-considered ones without burning out in the process. With the right tools, that’s a goal every leader can achieve.

Call to Action: Feeling inspired to lighten your decision-making load? Go ahead and try building your own Decision Assistant Agent in n8n, or whatever platform you prefer. Even a simple pros-and-cons generator using AI can make a world of difference. Start small, stay curious, and happy decision-making - with your new AI partner at your side!

AI thinking partner workflow

For the more technically curious, you can access the full n8n workflow JSON showing key nodes and their configuration in the post below:

Don’t worry if the technical details feel overwhelming - the key point is that all these steps can be automated. Once set up, the process from question to answer is seamless.

References

1. Decision Fatigue- Have You Fallen Prey To It?

2. How to get unstuck: tips for moving past analysis paralysis

3. Too Many Choices: How Decision Paralysis is Crippling Your Workplace

4. A Comprehensive Guide to Decision-Making Frameworks

5. Why AI Isn’t a Tool-It’s a Thinking Partner

6. AI Automation on a Budget: Self-Host n8n and Save Hundreds of Dollars.

---

If You Only Remember This:

• Building an AI Decision Assistant requires no coding skills - tools like n8n make automation accessible to any leader

• Voice input through messaging apps eliminates friction when you’re exhausted and need help structuring thoughts

• Memory systems allow your AI partner to learn your leadership style and reference past decisions for better analysis

• The complete workflow (trigger, transcription, analysis, response, logging) creates a seamless experience from question to structured answer

• Start small with one decision type, then expand as you see the value in action

You now have both the vision and the blueprint for AI-augmented decision-making. The midnight manager from our story doesn’t have to choose between exhaustion and poor judgment anymore. Neither do you.

Start small: pick one recurring decision type you face and build a simple version of this assistant. As Jakub has shown, the technology isn’t the barrier-it’s taking that first step from overwhelmed to empowered.

What decision keeps you up at night? That’s where you start. That’s where this becomes real.

---

PS: Many subscribers get their Premium membership reimbursed through their company’s professional development $. Use this template to request yours.

Let’s Connect

I love connecting with people. Please use the following connect, collaborate, if you have an idea, or just want to engage further:

LinkedIn / Community Chat / Email / Medium