Roadmap

What's next.

The honest list of what we're working on, what we're designing, and what we're exploring. No ship dates we'd blow past — but if any of these unblock something for you, tell us and it moves up.

Status legend

Buildingactively in development this quarterDesigningspec / wireframe stage; build followsExploringopen question, would-build-with-validation

Customer replication panels

Building

Upload your customer list, your CRM export, or a transcript pile — Coldread builds a panel that thinks like the people already paying you.

Today the Americans panel reflects the U.S. population. The next leap is panels that reflect YOUR population: your existing customers, your trial sign-ups, your churned accounts. Drop in a CSV (or connect a CRM) and Coldread synthesizes a panel sized to the cohort distribution. Now "would my customers shift on this copy" is answerable, not directionally — segment-by-segment.

Data-source-augmented agents

Designing

Connect Notion, Slack, Linear, your help-center, your sales-call transcripts. Each agent reads its relevant context before reacting.

When a doctor agent reviews your medical-product copy today, they react like a doctor. With data sources connected, they react like a doctor who has read your documentation, your competitor's positioning, and last quarter's support tickets — surfacing the real objections your team has already heard, not just the ones the model can imagine. First-class integrations: Notion, Linear, Slack, Intercom, Gong, Salesforce.

Custom report templates

Designing

Brand-matched, audience-specific report formats. Save your team's house style once; every export ships in it.

Today exports come back as Markdown, JSON, or CSV — the raw substrate. The next layer: visual report templates with your logo, your color palette, your section ordering. Save a template once for the marketing-team review, another for client deliverables, another for the legal-comms sign-off. Run → click template → ship.

Deeper MCP toolset

Building

More tools the agent can call mid-conversation: A/B/C/D pairwise sweeps, audience-segment drilldowns, follow-up campaigns scheduled across runs.

The current 19-tool MCP surface lets an agent run, compare, and drill in. Next: multi-asset sweeps (5 headlines vs 1 audience in one pass), saved follow-up question packs that auto-fire after every run, persona-cluster discovery ("which 30 panelists are most predictive of overall score on this kind of asset?"), and persistent agent memory across runs so iteration cycles compound.

Live moderator chat

Exploring

Drop into a panel mid-run and ask a follow-up before all 510 reactions land. Steer the conversation in real time.

Right now you ask follow-ups after a run completes. Live moderator turns Coldread into a live focus group: as reactions stream in, you can pause, surface a clarifying question to a sub-segment, and re-thread. Same agent infrastructure, conversational orchestration on top.

Audience snapshots over time

Exploring

Re-run the same asset against the same panel quarterly. Watch how the audience drifts — you'll see the trend before your competitors do.

Marketing copy doesn't age in a vacuum — sentiment around "AI," "sustainability," "layoffs," or any cultural lever moves with the news cycle. Snapshot mode lets you re-run the same asset on a schedule and chart how the panel response shifts. Useful for evergreen pages, brand pillars, and recurring campaigns.

Want to bias the priority?

Two ways to push something up the list:

Enterprise customers (custom panels, dedicated capacity, SLA) get first call on prioritization. Contact us if that's the right framing for your team.

← back to Coldread