SolutionsFor innovation labs
Governed explorationat enterprise scale.
Corporate labs need speed and a clear record. LiteSurface pairs structured discovery with budgets, provider controls, versioned prompts, and a decision trail your stakeholders can inspect, then carries the winner to a pilot app inside the same controls.
- Workspace budgets and provider breakers
- Prompt versions with rollback
- Audit log across projects
The situation
Exploration is easy to startand hard to account for.
Labs are asked to find new growth while answering to risk, legal, and finance. AI tools promise speed, but few can show what they sent where, what it cost, or why a recommendation changed.
Spend without a ledger
Model usage spreads across teams and personal accounts, with no budget that actually stops a run before it overspends.
Opaque recommendations
Sponsors cannot see which evidence, prompt, or model produced a score, so they discount it or ask for a manual rewrite.
Controls bolted on late
Governance arrives after the pilot, slowing down the teams it was meant to protect and forcing work to be redone.
Pilots that leave the guardrails
The moment a concept becomes a prototype, the work moves to other tools and personal accounts, and spend, access, and history stop being recorded.
The workflow
Exploration withthe controls built in.
Policy comes first, so every run afterwards happens inside it, from the first research run to the pilot sponsors click through.
- Step 01
Set policy first
Admins choose which providers a workspace may use, set budgets that stop runs when spent, and review the active prompt versions before the first run.
Governance - Step 02
Map your capabilities
Build an atlas around your platforms and partners, with availability and freshness tracked, so concepts rely only on what you can actually use.
Capability Atlas - Step 03
Evaluate with provenance
Every score records its evaluators, prompt versions, cited claims, and gate results. Confidence and disagreement are reported separately.
Evaluation - Step 04
Report to sponsors
Portfolio decisions carry rationale and history. Validation briefs and PRDs give sponsors a readable summary with footnoted citations.
Portfolio - Step 05
Build the pilot
The Builder builds against confirmed screens in a sandbox. Each request has a spend cap, workspace budgets cover all model spend, and the agent sees variable names, never values.
Builder - Step 06
Show a working pilot
Deploy a preview with its own Postgres database at a shareable URL. Unclaimed previews are torn down automatically; the code can go to a repository you connect.
Ship
What matters most
The parts labsare asked about first.
- GovernanceBudgets, provider controls, audit.Budgets, provider circuit breakers, prompt versions with rollback, and an audit log, all in a separate admin console rather than mixed into the product.
- EvaluationTwo evaluators, one transparent score.Deterministic code computes every score from the stored outputs of two independent evaluators, so any result can be recomputed exactly and explained line by line.
- PortfolioShortlist, validate, build, or hold.Human overrides are recorded next to the AI recommendation, with rationale, user, and time. Neither overwrites the other.
- EvidenceSources and atomic, cited claims.Retrieved pages are treated as untrusted data, never instructions, and every claim keeps a locator back to its source.
- BuilderAn AI agent builds the real app, live.A spend cap on every request and a version with a diff for every change, so a pilot’s code has the same kind of trail as the evaluation that chose it.
- ShipA pull request, a ZIP, or a live preview.Each deploy preview runs on its own Postgres database. On Enterprise a preview stays up 168 hours unless claimed into your own accounts, then it is torn down.
A pilot the risk team can sign off on
An innovation lab at a financial services company wants to explore customer-facing tools built on new mobile capabilities, within a fixed quarterly model budget.
An illustrative scenario, not a customer story. Names and numbers are invented to show how the workflow behaves.
Setup
An admin sets the allowed providers, a monthly budget, and project exclusions for regulated advice. The active prompt versions are reviewed.
Discovery
Research and concept generation run across three projects. One concept fails the legal and safety gate and is set aside with its rationale.
Mid-quarter
A provider outage opens a circuit breaker. Evaluation continues in single-evaluator mode, and each affected result is flagged as partial.
Quarter review
Sponsors review the portfolio, the usage ledger, and the audit log. Two concepts move to validation with experiments and kill criteria.
Next quarter
The concept whose experiments held gets screens in Mockup Studio and a pilot in the Builder, under the same budget. Sponsors try it on a deploy preview with test data.
Portfolio
Illustrative data- 77 out of 100Savings goal nudgesStrong bandValidate
- 74 out of 100Receipt capture for expensesStrong bandValidate
- 62 out of 100Card controls widgetHold bandHold
- 68 out of 100Automated credit adviceBlocked by a hard gateGate failed
The result
The lab moves at the pace of a startup team. The quarterly review answers what was spent, what was sent to which provider, and why each concept landed where it did, and the pilot was built under the same budget.
In the project at the end
- Provider policy and budgets
- Three evaluated projects
- Decision record
- Screens and specs
- Pilot app
- Deploy preview
Which plan fits
Start with Enterprise.Change plans as you grow.
Enterprise adds single-tenant or self-hosted deployment, custom scorecard templates, managed provider keys, and a security questionnaire and DPA, on an annual agreement. SSO and SCIM are planned.
Solo
Free
during early access
One workspace, unlimited projects, idea to app.
- 400 image credits a month
- 60 build minutes a month
- No deploy previews
Team
Free
during early access
Shared workspaces, roles, and the admin console.
- 4,000 image credits a month
- 600 build minutes a month
- Deploy previews
EnterpriseBest fit
Custom
annual agreement
Dedicated deployment, managed keys, and procurement support.
- 20,000 image credits a month
- 3,000 build minutes a month
- Deploy previews
A lab can begin with a Team pilot in one workspace and move to Enterprise once the deployment model is agreed.
Explore faster,with the controls already in place.
Tell us about your deployment, identity, and review requirements, and we will shape an Enterprise agreement around them.
- Solo and Team are free during early access
- Explore a sample project first
- Bring your own OpenAI and Anthropic keys