What’s coming
Read this first
These are directions, not commitments. Everything on this page is under consideration or in development. Any of it may change shape, be reordered, be folded into something else, or be dropped. Nothing here is a delivery commitment, and no dates are promised.
You should not plan a program, a budget, a procurement, or a security review around anything on this page. Plan against what mDecide does today, which is described in What is mDecide and the core concepts pages.
You will not find a timing column here, and that is deliberate. Publishing dates for work that has not been committed would be worse than publishing nothing: it would give you a schedule to plan against that we have not agreed to hold. If something below is load-bearing for your program, raise it at mdecidesupport@metricsiq.ai — it can be discussed against the actual roadmap, where commitments live.
The groupings below are thematic. They say nothing about sequence or timing.
Starting without an integration project
Today, connecting a new system and framing a new decision is work the engagement team does with you. The direction is to move more of that into your own hands.
| Direction | What it would mean for you |
|---|---|
| Connect your own data source | Point mDecide at a system you already use through a guided flow — no integration project and no data-engineering ticket. |
| Ask a question of your data | Ask an analytical question in plain language and get an answer with the work shown: the thing you used to wait on a data scientist for. |
| Describe the decision you need | State a decision in plain language and have mDecide assemble the pack, the data, and the agents needed to start the loop. |
| A plain-language check on the data behind a decision | Before a decision runs, see in business terms what data was found, what is missing, and what was substituted. |
Trust and reuse
This group is about knowing how much weight to put on a recommendation, and about not doing the same work twice.
| Direction | What it would mean for you |
|---|---|
| “Decisions like this one” | Before you decide, see the most similar past decisions, what was chosen, and how those turned out. |
| Calibrated confidence | When the platform expresses confidence, see how well that confidence has held up against actual outcomes on comparable decisions, so you know when to lean in. |
| Assumption tracking with drift alerts | mDecide watches the assumptions behind a decision you already made and flags you when one has moved far enough to warrant a revisit. |
| Save a decision as a repeatable one | Work a decision once, then save the whole setup as a template your team reuses — without engineering involvement. |
| Set your own decision checkpoints | You decide which decisions route to a human before anything is acted on — for example, anything above a threshold you set, or anything touching a particular account. |
| What-if scenario sliders | Change an assumption and watch the recommendation update, in place of a sensitivity analysis requested by hand. |
Compounding across the portfolio
This group is about the platform getting better at your decisions the longer you run them.
| Direction | What it would mean for you |
|---|---|
| Override library | When a person overrides an agent, the reason is captured and learned from, so the same override does not have to be made again. |
| Proactive decision surfacing | Rather than waiting to be asked, mDecide flags the decisions worth making now: the accounts at risk, the quotes going stale. |
How this page changes
This page is reviewed on the same cadence as the rest of the public documentation. Items leave it in one of two ways: they ship, in which case they move into the pages that describe the product as it is; or they are dropped, in which case they disappear from this page without ceremony. An item appearing here for a long time is not evidence that it is close.
Next
- Adopting mDecide — how an engagement actually starts.
- FAQ — evaluation-stage questions.