Upload the mess.
Zero setup, zero data prep.
Spreadsheets, CSVs, Parquet, PDFs — or connect your data warehouse. No ETL, no data scientist: the agent reads it all, structures what it needs and gets to work.
PredictFlow Project intake
Weekly sales by product
Uploaded: 9 files (sales_historical.parquet)
Yes, roll upKeep sizes split
Write to PredictFlow…
+ AttachEnter to sendSend
PredictFlow Model selection
Backtest · 32 seriesPer series ▾
MODELFAMILYMAE
TimesFM-3foundation8.4
Chronos-2foundation9.1
PatchTSTtransformer10.6
Prophetdecomposition12.2
Seasonal naivebaseline15.8
Moving averagebaseline17.3
WINNER BY SERIES
TimesFM-318
Chronos-28
Prophet4
Seasonal naive2
WHY THIS ONE
Beat the seasonal baseline on 7 of 8 held-out weeks. Kept the June peak out of the trend.
PredictFlow Board · General
test aggregateEdit+ New panel
PredictFlow What-if · scenario
Forecast · scenarioModel · v2Adjusted · v3
ActualsModel v2Adjusted v3W38 · 4 → 4,4 +10%
PredictFlow Accuracy · v1–v6
AccuracyShownModelCompare
MEAN ERROR
6.8%
118 evidence points
IN UNITS
4,1
per series / week
BIAS
+1.2%
slightly high
By horizonerror grows with distance
w1w2w3w4w5w6w7w8w9
SERIESv5v6TREND
W26-0696-003-7147,4%5,9% ▾
W26-0665-001-1009,1%8,2% ▾
W26-0522-001-10012,6%13,4% ▴
An agent that explains itself.
Not a black box.
Ask it anything.
“Why is this product falling?” or “what if price goes up 8%?” Every answer shows its reasoning and comes from your own data, not a modelling ticket.
Sharper every cycle.
Every approval, correction and new batch of actuals feeds back into the agent. Your forecasts improve on their own, month after month.
Your data stays yours.
Run it on your own servers, next to the files or warehouse you already use. Your data never leaves your infrastructure.
Comes wherever you are.
Your sheets, databases and tools.
- Excel
- Google Sheets
- Anaplan
- PostgreSQL
- HubSpot
- Slack
- Stripe
- .csv
- Salesforce
- b.
- Gmail
- BigQuery
- MySQL
- Snowflake
- Shopify
- Teams
- Zoho CRM
Start with one planner.
Scale to the whole team.
Individual Contributor
For the analyst forecasting from files.
$89/ Per Month
- Up to 20 projects
- Upload Excel, CSV, PDF and docs
- Explained, versioned forecasts
- Accuracy scored as actuals arrive
- No contracts, cancel anytime
Team
For teams forecasting straight from live data.
$439/ Per Month
- Everything in Individual Contributor
- Unlimited projects
- Connects to your databases, CRMs and tools
- Warehouses like Snowflake and BigQuery
- No contracts, cancel anytime
Enterprise
For companies whose data can't leave the building.
- Everything in Team
- Self-hosted on your servers
- Data never leaves your infrastructure
- Custom pricing
Need a custom deal?Volume discounts and flexible terms available for bigger commitments.Send emailQuick chat
“We used to lose the first week of every month rebuilding the forecast in spreadsheets. Now it's waiting for us, with the reasoning behind every number and a score for how last month's forecast actually did.”
Laura Méndez
Head of Demand Planning at NorthNood Foods
Questions.
Answered.
Brokyl is a small team of data and forecasting engineers. We built PredictFlow because planning teams were spending more time rebuilding spreadsheets than deciding what to do with the numbers.
Every plan takes Excel, CSV, Parquet, PDF, Word, text, Markdown and JSON uploads. Team and Enterprise also connect directly to your databases and tools, such as PostgreSQL, MySQL, Snowflake, BigQuery, Salesforce, HubSpot and Shopify.
Upload your files and the agent profiles them, asks you a few questions in plain business language, then builds the forecast and an annotated chart. Most projects have their first explained forecast in the same session.
It writes and runs its own Python against your history, tests forecasting methods and keeps what performs best. Every choice is explained on the chart. Once a forecast is live, it never switches method by itself: a person has to ask for a rebuild.
When real values arrive, a deterministic script, not the AI, scores every forecast version against them. The agent then explains those numbers in plain language, so the thing that made the forecast never grades its own work.
Always a person on your team. Each forecast is a numbered version the agent proposes; someone approves or rejects it before it becomes the official number. The agent has no way to approve its own work.
Yes, on the Enterprise plan. PredictFlow ships as a containerised deployment that runs inside your own infrastructure, with every project isolated in its own containers, so your data never leaves your network.
No. Your files and forecasts are used only to build and explain your own forecasts. Each project lives in its own isolated workspace, and you can delete a project and all of its files at any time.
Both get the same agent and the same forecasts. Individual Contributor works from uploaded files and is limited to 20 projects. Team adds unlimited projects and direct connections to your databases, CRMs and warehouses. Enterprise adds self-hosting on your own servers.
Yes. Plans are billed monthly with no contract, so you can cancel whenever you like.
Individual and Team plans are paid monthly by card. Enterprise can also be paid by invoice.
Since you can cancel anytime, we don't refund partial months. If something didn't work as it should, email us and we'll make it right.
Read our blog to learn more about the process and benefits.