Features

One workspace to protect all that makes your business bolde.

Connect your existing systems. Find reliable answers, share work, and put AI to use across your business—with the models and hosting you choose.

Whole-program architecture

Connect your systems. Put information to work.

Bring your information together so your team can find answers, spot changes, and work together. Choose cloud, private, or customer-owned models, with hosting configured separately.

Download Architecture Brief (PDF)Download Architecture Brief (Portrait PDF)
Model Strategy
Hosting Environment
01 · Ingestion & The Data Morass

Your Information Morass

Unstructured notes, call audio, spreadsheets, SaaS silos, and legacy databases. APIs or not—Bolde ingests it all.

APIs or Not · We Ingest It All
Cloud APIsContinuous sync via REST & webhooks
Raw Dumps & FilesMessy Excel, PDFs, CSVs & SFTP drops
Database ReplicasRead-only SQL replicas & lake shares
Custom AdaptersTailored for on-prem & legacy tools
GranolaMeeting notes & key takeaways
Fireflies.aiVoice transcripts & meeting recaps
AirtableRelational tables, bases & workflows
Microsoft SharePointDocument libraries & company intranets
Microsoft OneDriveCloud files, workbooks & sync folders
Google WorkspaceDocs, Sheets & shared Drive folders
NotionCompany wikis, notes & project trackers
SlackChannels, threads, huddles & decisions
Microsoft TeamsChats, channels & meeting transcripts
SalesforceCRM, accounts & revenue pipelines
HubSpotInbound marketing, deals & tickets
RipplingEmployee directory, payroll & HR ops
WorkdayEnterprise HR, headcount & financial planning
SnowflakeData warehouse & analytics tables
DatabricksLakehouse delta tables & feature stores
Oracle NetSuiteERP, general ledger & billing records
Sage IntacctFinance, accounting & audit reports
QuickBooksInvoices, expenses & ledger records
StripePayments, subscriptions & payouts
ZendeskCustomer service tickets & resolutions
IntercomCustomer conversations & AI chat logs
LinearProduct roadmaps, issues & cycle specs
JiraSprint backlogs & engineering boards
ConfluenceArchitecture wikis & technical specs
GitHubRepositories, PR reviews & code releases
PostgreSQL & MySQLProduction database replicas & models
Raw Files & SFTP DumpsMessy Excel, CSV exports, PDFs & S3 drops
+ SFTP dropboxes · S3 / Azure Blob buckets · Intranet wikis · Scraped portals · Air-gapped shares
Configured Ingestion Pipeline

APIs or not: we connect, ingest, and index your entire information sprawl into the Lake with verified source links.

01 · Existing Boundary

bolde. Perimeter

Selected Deployment Boundary

The surrounding protected enclave around Bolde services, approved information, and connected applications.

02 · Existing Guard

bolde. Sentinel

Security & Operations Oversight
Network & DNS boundariesSelected threat alertsAvailability & backup monitoringAuthenticated operator controls

Enforces agreed deployment controls; does not replace customer identity or application permissions.

03 · Existing FoundationShipping Now

bolde. Lake

Connected, prepared, source-referenced information

  • Source checking: Every fact links to verified tables, rows, or document sections.
  • Continuous freshness: Keeps organized as connected sources change.
  • Access protection: Preserves customer boundaries; private records stay private.
Intelligence & Oversight SuiteBuilt directly on Lake foundation
04 · Existing
bolde. Analyst

Sourced analysis on your data. Answers business questions and cites the exact records.

05 · Existing
bolde. Scouts

Change checks and concise 7 a.m. briefs on data movement and threshold anomalies.

06 · Existing
bolde. Auditor

Reviewable records of queries, answers, citations, and operator actions.

07 · Target <1 Mo
bolde. Evergreen

Living documents, slides, and spreadsheets whose facts refresh from live Lake sources.

08 · ExistingShipping Now

bolde. Space

An encrypted shared workspace forms around related people and work, with access set for the Space and its individual items.

Encrypted spacesItem permissionsPeople + agents
Configured WorkflowsCloud & Local Inference

Applications & Workflows

Analyst & SpaceCustomer AppsInternal IT ToolsPermitted Agents

Frontier cloud and local private models run directly inside Bolde applications under customer agreement and audit controls.

Progressive OrderClear boundary between existing and later releases
09bolde. ConsolePhase II · Forthcoming

Control & review deck for Operator.

10bolde. OperatorPhase II · Next

Repeatable permitted actions inside rules.

11bolde. KeeperPhase II · Planned

Approved Outlook filing with review.

12bolde. TodayPhase II · Planned

Priorities and next steps from work.

13bolde. CustomPhase III+ · Scoped

Customer-owned model shaped by agreement.

14bolde. RollbaxPhase III+ · Roadmap

Planned reversal of supported actions.

Dimension A · Models

Independent Model Choice

Model strategy is chosen separately from where the platform is hosted.

Inside PerimeterLocal Inference

Private Models

Selected open-source or open-weight models (e.g., Llama, Mistral) run locally in the controlled environment.

No data leaves your boundary
Inside PerimeterCustomer-Owned

Sovereign Models

Customer-owned models. Bring an existing proprietary model or commission ASG to create and tailor one under an agreed scope.

Customer holds ownership & weights
Used Inside Bolde · EgressFrontier Providers

Frontier Cloud Models

Claude Opus 5.5GPT-6 AstraGPT-6.1 SolGemini 4 Argon

Used directly inside the Bolde application. In Analyst, Space, and Evergreen, your team prompts frontier models from the native Bolde UI. Bolde manages prompt minimization and audited API egress, checking all returned answers against Lake citations. No full-lake export.

Seamless in Bolde apps · Governed API tunnel
Outside Perimeter · Egress strictly governed per query
Dimension B · Hosting

Choose Your Infrastructure

Deploy where your sensitive records already live. Model ownership and hosting platform are configured independently.

ASG-ManagedUS-Only
SOC 2 Data Centers

Managed environment in audited U.S. facilities with infrastructure controls. All Bolde enclaves built to NIST 800-172 standards.

Customer-OwnedOn-Premise
Your Private Facility

Run within your enterprise hardware, air-gapped facility, or sovereign perimeter.

Customer CloudAzure
Your Azure Tenant

Dedicated deployment inside your Microsoft Azure subscription and selected region.

Customer CloudAWS
Your AWS VPC

Isolated deployment inside your Amazon Web Services account and selected region.

Existing (Shipping now): Perimeter, Sentinel, Lake, Analyst, Scouts, Auditor, Space.
Target <1 month (16 Nov 2026): Evergreen.
Phase II (Forthcoming): Console, Operator, Keeper, Today.
Later phases: Custom (scoped model), Rollbax (action reversal).

Software-Led Adoption

A practical path toward bolde. AI-native operations

bolde. AI and data-science specialists work alongside your business and engineering teams to identify high-leverage agentic workflows, shape them into verifiable use cases, and configure bolde. applications and connections around the work your organization chooses.

01 · Identify

Find the right work

Prioritize repeatable, valuable workflows across business and IT.

DeliverableWorkflow shortlist & value criteria
02 · Shape

Make it actionable

Define the information, review points, and approvals each workflow needs.

DeliverableCitation rules & review gates
03 · Configure

Fit bolde. to your stack

Configure bolde. apps and authorized connections around the work your team chooses.

DeliverableConnected Lake & operator boundaries
Guiding Principle

Start with one valuable workflow, verify its sources and citations, then expand toward AI-native operations on a platform your team can operate.

bolde. Perimeter

The deployment boundary around the work

What it does

  • Defines where Bolde services, approved data, and connected applications run for a deployment.
  • Keeps the Lake and Bolde applications inside the selected environment boundary.
  • Sets the boundary for approved connections to private, sovereign, or explicitly selected public model services.

What it doesn't do

  • Does not make every hosting provider or model route equivalent.
  • Does not guarantee one location when a deployment explicitly sends approved content to an external model service.
  • Does not itself grant users or agents access; permissions remain enforced by the relevant systems and APIs.

Where it lives

Deployment-specific. Bolde managed hosting is U.S.-only in data centers with SOC 2 controls. Customer, Azure, and AWS environments follow their selected region, provider controls, and configuration.

What it cites

Deployment documentation identifies the selected hosting location, connected services, and model routes.

What gets logged

Boundary crossings and configured service connections are designed to be visible to the deployment operator.

Existing deployment layer — availability depends on environment

bolde. Sentinel

Security and operations guardrails

What it does

  • Provides network boundary and DNS controls for supported deployments.
  • Surfaces threat alerts and supports availability and backup monitoring.
  • Supports operator authentication controls for Bolde services and workflows.
  • Connects configured services through approved interfaces; access policy depends on the integration and deployment.

What it doesn't do

  • Does not imply that every API has a uniform fine-grained access policy.
  • Does not replace identity, authorization, network, or application controls.
  • Does not make a public model route private; external processing requires an explicitly selected and governed deployment configuration.

Where it lives

An existing part of Bolde security and operations. Enabled controls and integrations depend on the deployment configuration.

What it cites

Deployment configuration identifies supported connections and enabled access controls.

What gets logged

Threat and service-health signals are available to operators according to monitoring configuration.

Existing security and operations layer — controls vary by deployment

bolde. Lake

One trusted home for your information

What it does

  • Brings approved information from your files, business systems, and public sources into one place for Bolde to use.
  • Connect the platforms you already use—from major business systems to your own custom software. Bolde can use existing connectors or build tailored connections as part of your deployment.
  • New connections use authorized APIs, exports, databases, or other approved interfaces, with access and scope agreed for the deployment.
  • Keeps that information organized and current as connected sources change.
  • Shows where each item came from, when it changed, and who can use it.
  • Provides trusted information to Analyst, Scouts, Auditor, and Space. Evergreen is planned to build on the same approved sources.

What it doesn't do

  • Does not connect to a source without your approval.
  • Does not make private information public when it is brought together.
  • Does not turn conflicting records into a single answer without showing their sources.

Where it lives

Inside your Bolde deployment, within the access boundaries you set for each source.

What it cites

Every item keeps a link to its source and the time it was collected or refreshed.

What gets logged

Source connections, refreshes, access changes, and use of information leave a reviewable record.

Phase I — in progress / shipping now

bolde. Analyst

Sourced analysis on your data

What it does

  • Answers business questions from your data and cites the sources.
  • Shows confidence, source links, and the next question worth asking.
  • Runs on a schedule or when you ask.
  • Explains changes with links back to the records behind them.

What it doesn't do

  • Does not make decisions for you.
  • Does not send queries or results to public LLM APIs.
  • Does not read data sources you did not configure.
  • Does not create numbers without a source record.

Where it lives

Inside your network, connected to Evergreen. Inference, retrieval, and storage stay in your VPC or on-prem footprint.

What it cites

Every figure traces to the source table, query, and row range that produced it.

What gets logged

Each run records the user, time, data range, output hash, and citations used.

Phase I — in progress / shipping now

bolde. Scouts

Short briefs on what changed

What it does

  • Watch configured data sources for what changed and what needs attention.
  • Produce a short brief at 7 a.m. or after major data movement.
  • Rank what moved, by how much, and why it may matter.
  • Flag anomalies against thresholds you set.

What it doesn't do

  • Do not pull from public web sources unless you configure them.
  • Do not trigger actions automatically.
  • Do not keep raw intermediate signals beyond the analysis window.
  • Do not send briefs to personal email or consumer platforms.

Where it lives

Inside your network, on your schedule. Job metadata and outputs stay in your boundary.

What it cites

Each brief item names the data source, metric, time window, and baseline.

What gets logged

Each run logs start time, sources checked, signals reviewed, anomalies found, and delivery confirmation.

Phase I — in progress / shipping now

bolde. Auditor

A record of every important action

What it does

  • Keeps a tamper-evident record of queries, answers, citations, approvals, and Operator actions.
  • Makes answers replayable and exportable.
  • Flags policy issues early.
  • Produces audit packages with query logs, citations, access history, and output hashes.

What it doesn't do

  • Does not send audit packages to external services.
  • Does not alter log entries after they are written.
  • Does not delete records without explicit approval.
  • Does not decide compliance. It provides evidence for your team.

Where it lives

On infrastructure you control, with no replication to Bolde systems.

What it cites

Every log entry references the product, data source, and output hash.

What gets logged

Auditor also logs who requested an audit package, what it included, and the export hash.

Phase I — in progress / shipping now

bolde. Evergreen

Create work that stays current

What it does

  • Will create and edit living documents, slides, and spreadsheets that can be refreshed as their sources change.
  • Will let teams work directly in each document, slide, or spreadsheet.
  • Will support sharing individual artifacts and collaborating in Bolde Space.
  • Will connect approved sources so facts, figures, and citations can be refreshed and reviewed.

What it doesn't do

  • Does not overwrite your edits when a source changes.
  • Does not read sources you have not approved.
  • Does not make the decision represented by an artifact.

Where it lives

Inside your Bolde deployment, alongside the sources you have approved.

What it cites

Source links connect claims, figures, and charts to the records and data ranges behind them.

What gets logged

Version history records human edits, agent suggestions, source refreshes, approvals, and sharing changes.

Planned release target — within one month

bolde. Space

A secure place to work together

What it does

  • Brings people, agents, and related Evergreen artifacts into one shared workspace.
  • Lets owners grant view or edit access to one artifact or the whole Space.
  • Gives collaborators a place to edit, comment, and review agent contributions together.
  • Keeps each artifact connected to its approved sources as the work changes.

What it doesn't do

  • Does not make an artifact public just because it belongs to a Space.
  • Does not give an invited agent access beyond the sources and artifacts approved for it.
  • Does not turn a suggested agent change into a human-approved edit automatically.

Where it lives

Inside your Bolde deployment, with access set by the Space and each artifact.

What it cites

Shared artifacts retain the source links and update history attached in Evergreen.

What gets logged

Activity history records membership, access changes, edits, suggestions, and approvals.

Phase I — in progress / shipping now

bolde. Console

The command deck for Operator

What it does

  • Gives you one place to direct Operator work and see what is running.
  • Shows the task, its sources, its actions, and where your approval is needed.
  • Lets you review results and trace what Operator did.

What it doesn't do

  • Does not run work outside the rules you set for Operator.
  • Does not make the final decision for you.

Where it lives

Planned as the control surface for Operator in your Bolde deployment.

What it cites

Each result will link back to the inputs and sources Operator used.

What gets logged

Planned to record instructions, approvals, actions, and outcomes with Auditor.

Phase II — forthcoming with Operator

bolde. Operator

Repeatable work inside your rules

What it does

  • Runs repeatable work: retrieval, synthesis, document prep, and analysis pipelines.
  • Logs every action to Auditor.
  • Stops for approval at policy boundaries.
  • Keeps repeatable work inside the approved process.

What it doesn't do

  • Does not act without policy and logging.
  • Does not send data outside your network at runtime.
  • Does not make regulated decisions for people.
  • Does not hide behind the word "agent." It names the work it performs.

Where it lives

Inside your VPC or on-prem footprint. State and outputs stay on your infrastructure.

What it cites

Every action links to the source data, output, and action taken.

What gets logged

Every action records the user, time, workload, inputs, outputs, and action taken.

Phase II — next

bolde. Keeper

An orderly inbox, without the busywork

What it does

  • Learns the Outlook folders your team already uses and suggests where each new message belongs.
  • Shows you the folder plan before it starts moving mail. Filing begins only after you approve it.
  • Files clear-cut messages into the right folder and puts uncertain ones in a review queue for you.
  • Lets you correct a misplaced message so future filing improves.

What it doesn't do

  • Does not send, reply to, forward, or delete your mail.
  • Does not move mail before you approve the folder plan.
  • Does not guess when it is unsure where a message belongs.

Where it lives

Planned as a Phase II Bolde product for approved Microsoft 365 mailboxes, controlled from Outlook.

What it cites

Each filing decision can be checked against the message and the folder choice behind it.

What gets logged

Keeps a reviewable record of approvals, moves, uncertain messages, and corrections.

Phase II — planned

bolde. Today

The few things that need you today

What it does

  • Brings together approved mail, calendar events, tasks, and project updates so you can see what needs attention.
  • Puts one recommended next move first, with a short reason and the source behind it.
  • Keeps open commitments visible and gives you a quick place to capture something new.
  • Prepares follow-ups and replies as drafts for you to review.

What it doesn't do

  • Does not send a message or change a meeting for you automatically.
  • Does not treat every alert as equally urgent.
  • Does not hide where a recommendation came from or pretend stale information is current.

Where it lives

Planned as a Phase II Bolde daily work view connected only to sources your organization approves.

What it cites

Each priority points back to the message, meeting, task, or project update that explains it.

What gets logged

Keeps a reviewable record of captured work, proposed actions, approvals, and changes to priorities.

Phase II — planned

bolde. Custom

Customer-owned models, by agreement

What it does

  • Supports a model your organization owns and brings into its deployment.
  • Under a separately agreed scope, ASG can create or tailor a model for your organization.
  • The model, ownership, runtime, data use, and deliverables are defined in that scope.

What it doesn't do

  • Does not promise training or transfer of general base-model weights.
  • Does not use your data or feedback for model training without explicit agreement.
  • Is not included in standard deployments unless selected and scoped.

Where it lives

The model runtime and hosting environment are selected separately for each deployment.

What it cites

Planned to retain agreed source citations and review records for scoped model work.

What gets logged

Planned to record which approved feedback changed behavior, when, and by whom.

Roadmap — Phase III+

bolde. Rollbax

Later: reversal for approved Operator work

What it does

  • Planned for a later phase.
  • Designed to reverse approved Operator actions while keeping the audit trail.

What it doesn't do

  • Not part of the MVP or Phase II pilot.
  • Not a backup tool or disaster recovery system.

Where it lives

Planned for infrastructure you control, paired with Operator and Auditor.

What it cites

Planned to link each reversal to the original action and the outcome.

What gets logged

Planned to record who reversed what, when, and why.

Roadmap — Phase III+
Existing productsPerimeter, Sentinel, Lake, Analyst, Scouts, Auditor, Space
Release target · within one monthEvergreen
Phase IIConsole, Operator, Keeper, Today
Model routesPrivate · sovereign · public by separate agreement
LaterCustom, Rollbax

Ready to see it on real work?

Bring a sensitive use case. We will show how Bolde connects approved data, applies deployment controls, and keeps answers tied to their sources.

Request early accessView roadmap