Answers about your own figures

Your numbers, explained in a minute.

Lodestar reads your organisation's code and data and answers questions about your own figures the way an analyst would — working from the real source, showing how it got there, and saying plainly when a number couldn't be checked. In about a minute, not three days.

30s median time to a finished answer, where a ticket to an engineer takes days
$0.37 the cost of a typical answer
19/20 real code questions answered correctly, none wrong
15 built-in ways to get answers: reports, watchers, Slack and tool integrations

What you get

Not a chatbot bolted onto your data

A generic AI assistantLodestar

Your code

Sees only what you paste into the chat

All of the organisation's code, ten or more repositories read together, kept current from GitHub

Your data, safely

Your data sent off to a model you don't control, or an assistant given the keys to your systems

Works in a locked, read-only workspace that holds no credentials

Answers you can check

A confident answer with no way to tell a real figure from a guess

Every answer shows how it was worked out, and says so plainly when a figure wasn't read from data

Who sees what

Everyone sees everything, or nobody gets access

Access set down to a single column, by role or team, with a way to ask for more and a manager to grant it

Roles

One level of access for everyone

Administrators, stewards, managers, askers and viewers, each with its own permissions. A viewer reads what's shared with them without needing to ask anything

Reports

Copy each answer into a document by hand

Reports that refresh on a schedule, check themselves, and go to your team or out as a PDF

Watchers

You have to remember to ask

A question asked for you on a schedule, and a message when it finds something

Your tools

Answers you then re-type into the tools you actually work in

GitHub, Google Docs and Sheets, HubSpot, Slack, Asana and Basecamp, with every change shown to you and waiting for your yes

Spend

A bill you only understand at the end of the month

A daily limit, the cost of work before you run it, alerts and an instant pause

Records

No record of what was asked or answered

A searchable audit trail, and content cleared on schedules the organisation sets

Running it

A data team to set it up and keep it running

A console for setup, health, people and spend, and a user manual built in

Performance

Faster, cheaper, and more often right

Measured on a customer's codebase against a general-purpose assistant answering the same questions. Press the button to watch a typical answer from each, at ten times speed.

Time to a finished answer

Median, the answer most people get

117.6 s

Lodestar29.6 s
Done
Generic AI117.6 s

And the slow ones, the answer that one in ten waits longer than:

Lodestar69 s
Generic AI398 s

Cost of a typical answer

$0.76Generic AI
$0.37Lodestar

About half, with a spending limit you set and the cost shown before long work runs.

Right on real code questions

Lodestar19 of 20
Generic AI16 of 20
CorrectPartly rightWrongNo answer

Based on a month of use on a customer instance, over the same questions.

Roadmap

Where Lodestar goes next

In Progress

  • From request to releaseAnyone can raise a request in a conversation. It gets triaged, planned and shipped, and whoever asked hears when it's live.
  • Lodestar in SlackAsk in a direct message or a channel, and get the same answer you'd get in Lodestar.
  • Agreed definitionsDefine a measure once, and every answer and report counts it the same way.

Soon

  • Usage and valueSee more about how people are using Lodestar: which teams use it and how often, the questions they ask most, which reports and watchers get read, and where answers fall short, so you can see where it's helping and where it could help more.
  • Watchers that read textWatchers follow notes, check-ins and messages as well as numbers, and tell you in plain words what's changed. That might be a shift in tone across a team, or one note that says someone needs help.

Future

  • Working togetherVersions, comment threads and @mentions on shared work.
  • Better reports and dashboardsUpdates and improvements to reports and dashboards, shaped by what the people using them tell us.
  • Triggered work and briefingsWork that starts when something happens, not only on a schedule, and a briefing waiting for you each morning.

PRICING

Start with a pilot and grow from there.

STEP 1 · PILOT
$10,000
for 90 days
  • Up to 20 people
  • We set it up with you
  • Half credited to your first year
STEP 3 · EVERYONE
$30
per extra person, per month
  • Priced so the whole team can ask
  • 50 people: $39,000 a year
  • 100 people: $57,000 a year
You use your own AI provider account and pay them directly. We add nothing on top.A daily spending limit you set keeps the bill predictable.

We show our working on the price, too.

Pick a team size to see every line of the first year’s bill.

YOUR FIRST YEARUSD
People who can ask
40
Pilot, 90 days$10,000
Your own copy, 25 people included$30,000
15 more people × $30 × 12 months$5,400
Half the pilot, credited−$5,000
AI usageBilled by your provider
Total $40,400
$84 per person, per month

Stop waiting three days for a one-minute answer.

Book a demo