How Decision Flow Turns AI From Chat Into Action

You ask ChatGPT which database to use for your new service. It gives you 2,000 words comparing PostgreSQL, MySQL, MongoDB, CockroachDB, and three others you didn't ask about. You're now more informed and less decided than when you started.

This is the core failure of AI-as-chatbot. More information doesn't produce better decisions. It produces decision fatigue. And the gap between "I got advice" and "I took action" is where most of AI's value quietly dies.

We built Decision Flow to close that gap.

What Decision Flow is

Decision Flow is AI-assisted decision making built directly into Mentiko's orchestration platform. It's not a chatbot you ask questions to. It's a structured process that takes you from "I have a problem" to "here are my work items" in under five minutes.

There are three modes:

Intake mode for quick, low-stakes decisions where you just need a recommendation and a rationale.
Guided mode -- the three-round wizard -- for decisions that affect your team, your architecture, or your roadmap.
Classic mode for a traditional dashboard view where you compare options side by side with full detail.

Guided mode is the one worth explaining.

The three-round guided mode

This is the feature that changes how your team makes decisions. Three rounds, each one narrowing the space from "everything is possible" to "here's exactly what to do."

Round 1: Binary preference questions

Decision Flow starts by figuring out what you actually care about. Not what you think you care about -- what you reveal through quick, forced choices.

The AI generates 5 to 7 binary questions presented in a swipe interface. Each one forces a tradeoff:

"Is speed more important than reliability?"
"Do you prefer managed services over self-hosted?"
"Is team familiarity more important than technical fit?"

Swipe right for yes, left for no. No "it depends," no paragraphs. Just signal.

This takes about 30 seconds. By the end, Decision Flow has a weighted preference profile that drives everything that follows. The constraint is the point -- it surfaces priorities your team might debate for hours in a meeting but resolves in half a minute when framed as direct tradeoffs.

Round 2: Tailored options with match scores

Using your preference profile, the AI generates 3 to 4 concrete options. Not a generic list -- options scored and filtered against what you just told it matters.

Each option shows:

A match score (e.g., "87% match to your priorities")
Pros and cons filtered for relevance. If you said speed matters more than cost, the cost section is secondary. If you said team familiarity is critical, each option shows your team's experience level with that technology.
A risk summary -- what could go wrong if you pick this option, specific to your context.

The options aren't static. You can tap into any option to drill deeper, ask the AI to generate a variation, or adjust your Round 1 preferences and watch the scores recalculate in real time.

Round 3: Execution plan

This is where Decision Flow diverges from every other AI tool.

You select your choice. Decision Flow generates a full execution plan: a task tree with dependencies, time estimates, and assignees suggested based on your team's skills and availability.

The task tree isn't a document. It's structured data. One click and every task becomes a bead issue in your project tracker -- with dependencies wired, estimates attached, and the decision context linked so anyone can trace why this work exists.

The decision doesn't end as a Slack message or a Google Doc. It becomes work.

Real example: "Which database for our new service?"

Let's walk through a complete flow.

Your team is building a new notification service. You need a database. You open Decision Flow and type: "Which database should we use for our notification service? We expect 50k events per hour, need 99.9% uptime, and the team is mostly Python developers."

Round 1 fires. You swipe through seven questions. You reveal that operational simplicity beats raw performance, that managed services are preferred, and that you'd rather over-provision than optimize later.

Round 2 generates three options:

Amazon RDS PostgreSQL -- 92% match. Managed, familiar to Python teams, handles your throughput easily. Risk: vendor lock-in on AWS-specific extensions.
PlanetScale (MySQL) -- 78% match. Strong scaling story, but your team has less MySQL experience. Risk: learning curve slows initial development.
Supabase PostgreSQL -- 85% match. Open-source friendly, great developer experience. Risk: smaller operational track record at scale.

You pick RDS PostgreSQL.

Round 3 generates the execution plan:

Provision RDS instance with read replica (2 days, assigned to your infra lead)
Set up connection pooling with PgBouncer (1 day, depends on provisioning)
Create schema migration pipeline (1 day, assigned to backend dev)
Write integration tests against staging instance (2 days, depends on schema)
Update runbook with failover procedures (1 day, parallel track)

One click. Five bead issues created, dependencies linked, sprint-ready.

Why this matters

Decisions are the highest-leverage activity in any organization. A bad database choice costs you six months. A good one compounds for years. But most teams make critical decisions in Slack threads, on Zoom calls with no notes, or by deferring to whoever talks loudest.

Decision Flow changes the unit of output. The result of a decision isn't a recommendation -- it's an execution plan. The AI doesn't just help you think. It helps you commit and then hands you the work items to follow through.

And because every decision is stored with its context -- the preferences, the options considered, the rationale for the final choice -- you get something most teams never have: a decision retrospective. Three months later, you can review the decision, compare predicted outcomes to actual outcomes, and feed that data back into future decisions.

Bad decisions compound. Good decisions compound. Decision Flow makes sure you're compounding the right ones.

What's coming next

Decision Flow is shipping with guided mode in the current release. Here's what's on the roadmap:

Multi-stakeholder voting. Instead of one person swiping through Round 1, your whole team does. Decision Flow aggregates preferences, highlights alignment and conflict, and generates options that account for everyone's priorities.
Decision templates. Common decisions -- tech stack selection, vendor evaluation, hire/no-hire, feature prioritization -- come pre-structured with domain-specific questions and evaluation criteria.
Chain integration. A decision can trigger a Mentiko chain. Pick your database, and the provisioning chain runs automatically. The decision doesn't just produce work items -- it executes them.

Decisions shouldn't end as documents. They should end as action. That's what Decision Flow does.

Join the waitlist to get early access.