VERIFIED ANALYTICS AND REPORTING
AI-assisted, never AI-decided: models propose, and code proves every number and statement and traces it to its source before anyone sees it.
Analytics, reporting and document control for regulated organizations. It reads data where it already lives and returns finished reports, quick answers and searchable evidence, all checked, traced and released under the source’s rules.
Runs as a managed service in the United States, or on your own infrastructure: your cloud or data center, a disconnected install, a government enclave or the tactical edge.
Models draft the plan, the SQL and the prose. Deterministic code decides what runs and what is printed, and people approve wherever your organization’s rules ask for it.
01
In the web app, through the API, or from an AI assistant. One question, or up to eight in one report.
02
You see which data it will read and how, before anything runs, and can edit it.
03
Code runs read-only, checked queries. The writing model never types a number itself.
04
Each number is computed again a second way, and each sentence is checked against the values it cites. A mismatch stops the report.
05
Files in every format, each Office file and PDF signed, released only to the people your rules allow, with a record of each copy.
Your rules decide who approves a plan, who signs a report and who may release it. Every delivered file prints which of those happened, so no one has to take it on trust.
Examples of the line printed in each file
| Step | What a model is given |
|---|---|
| Planning | Table and column names, descriptions, masked samples |
| Writing | Labels and value ids, never your rows |
| Checking claims | Each sentence and the values it cites |
| Computing | Nothing. Numbers come from code |
Health data goes only to models covered by a business associate agreement, through an input and output screen that fails closed.
Business intelligence tools show what a query returned. DataLogiX shows how each number was reached, proves it again, and controls where the report goes next.
Every number is computed a second, independent way, or held for a person’s decision, and read back from every finished file. A second model checks each sentence against its evidence, and code refuses what the evidence doesn’t support.
Put a report before a board, a regulator or a court and show why each figure and sentence is right.
Every sentence, number, chart, map and marking traces to its origin in a signed record, and every Office file and PDF ends with an Evidence appendix printed from it.
Anyone can answer “where did this come from?” for any sentence or cell, in any file.
Releases, bound links, numbered copies, receipts, recalls and corrections, each hand-off in a signed record. Markings are read and written in US CUI, NATO, UK and Traffic Light Protocol schemes, and files leave encrypted to the recipient’s key or through your own key service.
Share at home and abroad, prove who received what and why, and call it back.
Databases, warehouses, documents, scans, mail, audio and images, read where they live. Data is inspected for what you must not hold, and the provider’s markings, purposes and terms travel with it.
Connect what you have, including merged companies’ and partners’ data. No migration first.
Find a person, organization, vehicle, vessel, account or identifier across every document, record and image, with ranked links, places and timelines.
Every search has a purpose and a reason, is logged before it runs, and respects each reader’s clearance and need to know.
Your sector’s definitions, writing standards, markings, records and release rules come as signed, versioned packs. A new version takes effect only when a second administrator approves it, and every run records the versions it used.
Your sector’s rules, applied the same way every time, and no one can quietly loosen them.
The web app, two SDKs, webhooks, event streams and an MCP server all use one public API. Assistants act only within the signed-in person’s rights.
Keep your own tools and assistants; they call DataLogiX under the same rules.
Each organization has its own cloud project and keys, and each client its own wall and key. Every action is audited, and the audit log can stream to your SIEM.
Shorter security reviews, and an on-premises relay for data that stays on your network.
The web report, workbook, deck, Word document and PDF are each read back after rendering and compared with the values on record before anyone receives them.
Every Office file and PDF is signed. Anyone can check one on the public Verify a file page, with or without an account.
| File | Read back | Signed | SHA-256 |
|---|---|---|---|
| Web report | Matches | Not a file | 9f2c41…a07e |
| Excel workbook | Matches | Yes | 31be08…5d12 |
| PowerPoint deck | Matches | Yes | c7d920…e4b9 |
| Word document | Matches | Yes | 5a08f3…71c4 |
| Matches | Yes | e612bd…0f3a |
Sensitivity labels and markings on incoming files are read and kept. Files above your organization’s ceiling wait for a person to decide.
Snowflake, Databricks, Redshift and other cloud warehouses. SQL Server, PostgreSQL, MySQL and Oracle through an on-premises relay. Document databases and vector stores.
Spreadsheets, PDFs and scans, Word and older Office files, email with attachments at any depth, archives, recordings and images.
HL7 v2, FHIR, C-CDA, X12 claims, CMS assessments. Ledgers, SAF-T, XBRL, e-invoices and EDI, bank statements.
Salesforce, accounting systems, Microsoft 365 and Google Drive, SFTP and watched folders, Amazon S3, REST APIs, security telemetry.
Password-protected files open once you supply the password, for one file or as a standing rule for a source. OpenPGP, S/MIME, ZTDF and Purview files open with your keys, and signatures are checked against the senders you register.
Truncated PDFs, broken archives and damaged Office, text, JSON and XML files are repaired automatically, and what was lost is counted. A PDF that can’t be repaired is read page by page from its images. Hostile files are held, with the reason.
Workbooks with merged headers, subtotals and pivots are read only when every total adds up. A text layout never seen before gets a parser drafted, checked by code and confirmed by a person before anything loads.
Data is inspected for what you must not hold: health data, identifiers, secrets, prohibited terms and known leaked records. It is held or redacted by your rules.
Data from merged companies and partners is matched across systems without an alignment project.
Markings, permitted purposes, usage terms and allowed AI models follow the data into every report.
ANALYSIS TO EXPLORE
The same proof runs at every size, from a quick answer to a dashboard.
One verified number in seconds, with the query that produced it a click away.
Two to eight questions in one report, six by default, each planned, checked and traced.
Dashboards of released reports, and maps by country and state in every file. Click a bar or a map area to see what lies beneath it.
Forecasts with intervals, regression, difference-in-differences, survival analysis and clustering. Words such as “trend” or “significant” need a test behind them.
LIVE DATA
Live data doesn’t skip the checks: each item is verified as it arrives, and a picture built from sensors still needs a person or a second sensor before a detection counts.
Sensor data comes through SensorLogiX: edge collectors, any link, a forward node that keeps working cut off, image and video search, and detection on your own GPUs. Civil and commercial uses
SUBJECT SEARCH AND DATA MINING
People, organizations, vehicles, vessels, accounts and identifiers, found across documents, records and images, with the evidence for every link.
A person, organization, vehicle, vessel, account or identifier, found across every document, record and image.
Links, places and timelines show the strongest connections ranked, with typed relationships and alerts on new or unusual links.
Names and numbers in the text of images are found with the box they sit in. Whole records such as incidents, tickets and transactions are searched by their fields, and images by what they show, by a similar picture and by time and place.
Every search needs a purpose and a reason, is logged before it runs, and can need a second person. Results respect clearance, compartments and need to know.
EXCHANGE ACROSS STANDARDS AND BORDERS
Regulators, authorities, NATO and allied partners and customers, at home and abroad. Every hand-off is approved, recorded and recallable.
Approved before they go, by two people where the rules say, and only to readers whose clearance, compartments and need to know allow it. Links are tied to one person, copies are numbered and fingerprinted, and receipts, recalls and corrections are recorded.
Every hand-off gets a signed record, and health data keeps HIPAA’s accounting of disclosures.
Markings read and written in US CUI, NATO, UK and Traffic Light Protocol schemes, and as Microsoft Purview labels. Once your marking pack maps them, IC-ISM and STANAG 4774 labels on incoming files are read, and export jobs write NATO ADatP-4778.2 bindings and IC-ISM sidecars beside each file.
Files leave encrypted to the recipient’s OpenPGP key, or through your own key service: an OpenTDF platform wraps them as ZTDF, the Zero Trust Data Format allied partners use for data-centric sharing, or Microsoft Purview protects them. The key service decides who opens each file even after it leaves, and its access records come back to you. Key-service wrapping is turned on by your organization’s policy.
Sensitive parts are removed, and each version is proven to keep every judgment and number unchanged.
Transmittal records and traces can carry an ML-DSA-65 signature beside ECDSA, turned on by policy. Connections to the web app, the API, the relay and SFTP use hybrid X25519 and ML-KEM-768 key exchange where the other side supports it, so traffic recorded today resists decryption by a future quantum computer.
Connect an OpenTDF platform or Microsoft Purview, and require wrapping from a marking level up. Your OpenTDF platform is also asked before a report page shows its content. ZTDF and Purview files sent to you open under your own grant, or are held until you grant it.
A whole collection leaves as one job: a signed manifest of every file, a second person’s approval above a size, encryption to the receiver, IC-ISM or ADatP-4778.2 labels where your marking pack maps them, and a receipt when it lands.
Access requests under HIPAA, GDPR and CCPA are tracked to their deadlines, the person’s identity is proven, and the answer is assembled from what you hold. Legal holds keep anything they name from being deleted.
Payment formulas, disclosure limits, analytic standards, markings, release rules and report types ship as versioned, signed packs. Experts can read and test them, and an organization or one of its clients adds its own on top.
Seven layers merge in order and each run freezes the result, so a report can always say which rules made it. Review gates, release tiers and budgets can only get stricter as layers are added.
CHOOSE A LAYER
LAYER 1 OF 7 · WHAT EVERY RUN STARTS FROM
The safety floor DataLogiX sets for everyone: the checks every query passes, the per-run spend cap, and plan review for an organization’s first runs.
Fixed by EVAS
LAYER 2 OF 7 · A RECOGNIZED STANDARD
A published standard, encoded as a pack and tested against the document it comes from: analytic tradecraft, records rules, marking systems, likelihood language.
Each rule cites its source
LAYER 3 OF 7 · YOUR SECTOR’S RULES
The rules of one sector as data and tested functions: payment formulas, rule tables, small-cell suppression and reliability thresholds, with the sector’s report types.
Versioned and signed
LAYER 4 OF 7 · HOW YOU WORK
Your settings on top: who reviews plans, how strict release is, your templates, house style and the people who approve and sign.
Can only tighten
LAYER 5 OF 7 · ONE CLIENT’S OWN TERMS
For organizations that serve clients: a client’s markings, release rules and template, kept to that client’s work alone.
Can only tighten
LAYER 6 OF 7 · A KIND OF REPORT
A report your sector repeats, such as a governing-body quality review: its sections, the analysis steps behind each one, and the review chain that signs it.
Signed before release
LAYER 7 OF 7 · THIS REPORT
The run itself. The seven layers merge in order and the result is frozen with the run, so the report can always say which rules made it.
Frozen per run
QAPI governing-body packs, value-based purchasing reviews, hospice cap monitoring, PEPPER, payer contracts
CMS rule tables · small-cell suppression · NCHS reliability · HIPAA records and accounting
Month-end close, receivables, bank reconciliation, two systems reconciled, statutory filings
General ledgers · SAF-T and XBRL · e-invoices and EDI · bank statements
Intelligence assessments, release to partners, link analysis, lower-marked versions
ICD 203 and 206 · NATO AJP-2.1 · UK PHIA · CUI and NATO markings · two-person release
Criminal intelligence products, threat briefs, subject search with purpose and oversight
28 CFR Part 23 · LEAS · Europol 4x4 · purpose and accounting per search
Monthly security reports from Defender, Entra, Sentinel and AWS, client threat briefs
FIRST TLP 2.0 · STIX confidence · MITRE ATT&CK techniques · Canadian Centre for Cyber Security
Benchmarking across members, evidence reviews, statistical bulletins
IPCC uncertainty · GRADE certainty · nine style guides · community release markings
A four-archetype suite runs whole pack stacks end to end before release: an intelligence cell, an OSINT firm, a hospice and a community benchmarking group.
Field operations come in from CareLogiX. CareLogiX’s visit and mileage exports are read like any other source, so DataLogiX can report on them under your rules.
How they connectSensor data comes in from SensorLogiX. SensorLogiX adds sensing, image and video search, and analysis on your own GPUs. Video, imagery, radar and tracks arrive confirmed and marked, over any link, for DataLogiX to turn into verified reports and controlled releases.
See SensorLogiXPeople use the web app. Your systems use the TypeScript or Python SDK. AI assistants connect through the MCP server. All three follow exactly the same rules.
Run by us in the United States, with a separate cloud project and keys for each organization inside a security perimeter, and recovery copies in a second US region.
The same platform on your own infrastructure, with your own sign-in and your own key store.
A signed offline install with local AI models and local prompt screening. It never calls home.
Smart-card sign-in and FIPS-approved algorithms only. Classified work runs on your accredited infrastructure.
With SensorLogiX: collectors beside each sensor, and a forward node on rugged servers that works disconnected with smart-card sign-in and syncs with the rear when a link returns.
Under every option, an outbound-only relay reads databases and file shares that stay on your network, with no inbound firewall port.
HOW IT COMPARES
Operating platforms, BI copilots and warehouse assistants answer questions and explain how. DataLogiX produces numbers it has proved, controls every copy that leaves, and lets their assistants call it as the verified tool.
| Capability | DataLogiX | Palantir Foundry and AIP | BI copilotsPower BI, Tableau, ThoughtSpot, Looker | Data platforms’ AIDatabricks, Snowflake, Google Cloud | Microsoft 365 Copilot and Purview |
|---|---|---|---|---|---|
| PROVING THE NUMBERS | |||||
| Answers questions over your data | Quick answers, and reports of up to eight questions | AIP Analyst | Chat and search over semantic models | Genie, Cortex Analyst, data agents | Copilot in Excel, Analyst agent |
| Every number computed a second, independent way | Before anyone sees it | Verified answers and queries approved by people | Benchmarks and verified queries | Users are asked to verify | |
| Every sentence checked against its evidence | Unsupported sentences are refused | Citations shown | Users are asked to review | Claims scored, not blocked | A review step; nothing blocked |
| Numbers read back from every delivered file | Web, Excel, PowerPoint, Word, PDF | ||||
| Dashboards and drill-down | Tiles from released reports, still verified | Contour and Quiver; chart drill-down | Drill-down, drill-through, actions | Databricks dashboards, drill-through | PivotTable and Power BI drill-down |
| Lineage for each value | Per value and per element, signed | Data lineage | Column-level lineage | ||
| Analytic standards such as ICD 203 | Likelihood and confidence computed by code | ||||
| Statistical methods built in | Survival, logistic, clustering; each run twice | Regression lines, forecasts; the rest in code | Regression, k-means, forecasts; R or Python beyond | Regression, k-means, forecasts; no survival | Excel statistics; deeper via Python code |
| Competing hypotheses, source grading, calibration | Computed by code, scored against outcomes | Sources cited; no grading | |||
| READING WHAT YOU HOLD | |||||
| Documents, scans, email and recordings as data | Read where they live, damaged files repaired | Documents, email, audio and video | Documents through add-ons | Documents, scans, audio, video; not email | Files and email; recordings by transcript |
| Search inside images and video | By what they show, similar pictures, time and place | OCR and embeddings as pipeline steps | Image and video embeddings in SQL | Text in images; video by transcript | |
| Live streams and sensor video | Each item verified on arrival; video with SensorLogiX | Streams and imagery; video in batch | Streams; imagery not found | Streams; satellite imagery in one platform | Live meeting transcripts |
| Subject search and link analysis under oversight | Purpose, reason and second person, logged first | Purpose recorded; approval not found | Graph links; gated search not found | Graphs and logs; purpose not found | eDiscovery cases and audit |
| CONTROLLING EVERY COPY | |||||
| Markings computed per paragraph | Portion marks and banners in every format | Markings travel with the data | Purview labels on some exports | Tags and policy tags | Labels; output inherits the highest |
| Clearances, compartments and need to know | Per reader and per item, before anything shows | Classification markings per resource | Row-level security; no clearances | Tag-based policies; no clearances | Labels per user or group; no clearances |
| Release chains, numbered copies, receipts, recall | Two-person release, originator consent | Purpose checkpoints on export | Track and revoke for Office files | ||
| Control after a file leaves | Recipient’s key, ZTDF or Purview | Encrypted values, not whole files | Labels on some Power BI exports | Revoke inside the platform only | Label encryption, revoke for Office files |
| Signed outputs | PAdES and C2PA | Images made with Designer | |||
| Post-quantum signatures and connections | ML-DSA co-signatures, ML-KEM key exchange | Vendor roadmaps; product not named | One vendor’s program, not product-specific | Quantum Safe Program; 2029 target | |
| HIPAA accounting of disclosures | Per person, as a product feature | ||||
| WORKING WITH WHAT YOU HAVE | |||||
| AI assistants over MCP | MCP server | Palantir MCP | Servers documented | Servers documented | Uses MCP servers as tools |
| Reads warehouses in place | Cloud warehouses, and databases on your network | Through its own platform | Through their models | Their own | Analyst agent over databases |
| Runs disconnected or on your own infrastructure | Your data center, an enclave or the tactical edge | Air-gapped, classified and edge | Self-hosted servers; offline copilot not found | One vendor’s air-gapped cloud only | Government clouds; offline Copilot not found |
| WHERE OTHERS ARE AHEAD | |||||
| Drag-and-drop, self-service exploration | Quick answers instead of drag-and-drop | Workshop, Contour, Quiver and Vertex | Built for it | Notebooks and dashboards | Excel and Power BI |
Competitor columns state only what each vendor’s public documentation said when it was read, between 29 October and 2 November 2026, summarized across the products in each group. “Not found” is never a claim that a product lacks a feature. DataLogiX doesn’t check numbers another tool made; it runs alongside these tools and reads their warehouses in place.
Investigation and link-analysis tools
i2 Analyst’s Notebook, entity-resolution engines, public-records services
DataLogiX searches your own data under purpose, second-person and need-to-know rules, and every link it shows carries signed evidence.
Motion-imagery, imagery and maritime tools
FMV players, imagery viewers, AIS and radar displays
DataLogiX takes their standard formats and feeds, and turns what they find into verified, marked, released products, with SensorLogiX at the edge.
FOR DEVELOPERS AND INTEGRATORS
Every screen in the DataLogiX app runs on the same public API you get: more than 540 operations, TypeScript and Python SDKs, signed webhooks and an MCP server for AI assistants. Build it into your own product, or run it for your clients from your own code.