Telepatia

Built AI native from launch in July 2025 rather than adding models to existing software, which the company and its investors frame as the advantage of building in a market without decades of accumulated health IT. Documentation, decision support, and error detection are all inference; there is no legacy product underneath.

The position is stated explicitly and is notable given the company markets a suite it describes as AI healthcare employees including AI doctors and nurses: it supports clinicians rather than making decisions, with the physician always making the final call and retaining clinical responsibility. Error flagging in real time so the clinician can adjust course is decision support by design, since the system surfaces a concern rather than acting on it. Holding a clear ceiling while using employee framing in marketing is a distinction buyers should note, and here the ceiling is documented.

No model provider, architecture, version or evaluation methodology was located, and nothing states whether an external foundation model service processes encounters.

What is disclosed is the training corpus, and it is more specific than most of this category offers. The company states its system is trained on peer reviewed literature, clinical guidelines, and local institutional protocols in the region it serves. The last of those is the distinctive claim and deserves examining rather than accepting.

Grounding a clinical system in each institution's own protocols is a genuine differentiator. It means recommendations reflect what that hospital has decided rather than a generic standard, which matters more in markets where international guidelines are adapted to local resource constraints. It also creates obligations nobody has described. Establish how institutional protocols are ingested, who validates that they were captured correctly, how they are kept current when the institution revises them, and what the system does when a local protocol conflicts with a published guideline. A system trained on a superseded protocol will continue recommending it confidently.

A second question follows from the same design. Where protocols from one institution inform a model serving others, a buyer should establish whether their own protocols remain their own, or whether they contribute to a shared model that competitors also use.

The personalisation claim, that output adapts to each clinician's style, raises the ordinary version of the same question about what is learned and retained per user.

Ask which foundation models are used, how the protocol corpus is maintained, and whether institutional content is segregated.

The corpus is described more specifically than most of this category manages and the operating chain is not described at all. The company states its system is trained on peer reviewed literature, clinical guidelines and local institutional protocols in the region it serves, and the last of those is the distinctive element.

Institutional protocols are the customer's own material entering model development, which is a form of the cross client question this index tracks everywhere, and it arrives in an unusually consequential shape: a hospital's internal clinical decisions are among the most sensitive non patient content it holds.

A buyer should establish whether its own protocols remain segregated to a model serving only it, or whether they contribute to a shared model that other institutions, including competitors in the same city, also use. Nothing public answers that. On enumeration the record is bare.

No foundation model provider, model class or version is named, no hosting arrangement is published, no sub processor list was located, and nothing states whether an external model service processes encounters at all, which for a platform spanning documentation, a clinical copilot, an interoperability layer and analytics means several different classes of content are moving to unnamed parties. Ask which foundation models are used, whether institutional protocol content is segregated per customer, and how the protocol corpus is ingested and maintained.

Unusually specific for a company one year old, with named sites and figures: protocol adherence reported rising from 84 percent to 99 percent, 1.7 to 2 hours recovered per physician daily, 60,000 errors flagged in real time, over 5 million consultations processed, and Mater Dei reporting roughly eight hours of daily use per physician. Sustained daily use at that intensity is a strong adoption signal because it is hard to fake.

Held back from A because all figures are vendor or investor stated without published methodology, and the errors prevented count in particular needs a definition of what was counted and how prevention was established.

No retention schedule, deletion commitment, de identification practice or statement on whether customer content trains models was located. What is published in their place is a claim to the highest standards of clinical data protection and management, which names no standard and belongs with the constructions this index treats as marketing rather than assurance.

The platform's breadth makes the omissions wider than they would be for a scribe. Alongside documentation it operates a clinical copilot, an interoperability layer moving information between systems, an analytics product turning clinical activity into performance measurement, and a stated suite of AI healthcare workers including auditors. Each holds a different class of content, and the analytics and audit functions imply retention of clinical activity data over time rather than transient processing.

The training question is the sharpest one on this record and it connects directly to the vendor's stated approach. The system is described as trained on local institutional protocols, which means institutional content already flows into model development. A hospital should establish whether its own protocols, its clinical activity data, or its encounter content contribute to models serving other customers, and whether anything in the contract prevents it.

Brazil's data protection law treats health information as sensitive personal data with a defined legal basis required for each processing purpose, so these questions have concrete legal answers the vendor should be able to supply.

Ask for the retention schedule per product, the legal basis relied on for each processing purpose, and the training position in contract language.

This axis does not read as an absence. The vendor is headquartered in Brazil and deployed across hospital systems in Brazil, Colombia and Mexico, with stated expansion plans toward other emerging markets rather than the United States. The United States health privacy rule is not the governing regime, and its absence is expected rather than a gap.

What governs instead is a set of national data protection laws that a buyer should assess on their own terms rather than by analogy. Brazil's general data protection law treats health information as sensitive personal data, requiring a specific legal basis for processing and imposing conditions on international transfer. Colombia and Mexico each operate their own personal data regimes with their own consent and transfer rules. None of them maps cleanly onto the United States framework, and a vendor operating across all three is subject to differing obligations in each.

So the practical instruction differs from the usual one. Rather than asking for a business associate agreement, establish which national regime governs your deployment, what legal basis the vendor relies on for processing health data, and what the data processing agreement says about transfer between the three markets.

One caution for any buyer outside the region. Nothing located establishes that a United States compliant arrangement is available at all. A vendor operating solely under other regimes may have no mechanism to offer one, and using such a tool on protected health information without it is a violation regardless of how well built the product is.

No named or dated attestation, no report of either type and no trust centre were located.

The expectation is raised by scale rather than by company size. The platform is deployed across more than twenty five hospital systems in three countries, is used for hours a day by physicians at named institutions, and has raised substantial venture funding. Enterprise hospital deployment at that level normally involves a security assessment by each institution, so private evidence exists in the hands of customers; none of it is published.

The relevant benchmark for these markets is worth naming, because a buyer used to United States frameworks may look for the wrong thing. There is no regional equivalent to the sector specific certification schemes found elsewhere, so the international information security management standard is the usual anchor, sometimes alongside a service organisation control report where a vendor sells internationally. Either would be a meaningful signal here and neither was located.

Two features of the platform enlarge what an examination would need to cover. The interoperability layer moves information between hospital systems, which means the vendor holds a path into institutional records rather than only receiving encounter audio. And the analytics product retains clinical activity data over time for performance measurement.

Ask which report or certification is held or scheduled, what boundary it covers, and specifically whether the interoperability layer and the analytics store are inside that boundary rather than only the documentation service.

Graded against the regime that actually governs this vendor rather than the United States device pathway. The row previously sat under a non canonical axis name and rendered nowhere; the content is unchanged and it now occupies the correct slot.

Operating ahead of settled regulation, which the company acknowledges. Brazil's Senate has approved a risk based artificial intelligence framework pending lower house and presidential approval, and Colombia has sent its own bill to Congress, so the compliance environment for clinical AI in its primary markets is still forming.

That acknowledgement is worth crediting rather than treating as a gap. A vendor stating plainly that the rules in its markets are unsettled is more useful to a buyer than one asserting compliance with a framework that does not yet exist, and this index has repeatedly found the opposite behaviour elsewhere.

The practical consequence is that obligations are a moving target. A framework adopted after deployment can impose classification, evaluation or transparency duties retroactively on systems already in clinical use, and the buyer rather than the vendor typically absorbs the remediation.

One feature of the platform sharpens that exposure. Alongside documentation the vendor operates a clinical copilot delivering evidence based recommendations and institutional protocol guidance at the point of care. Under a risk based framework of the kind Brazil is considering, systems influencing clinical decisions are the ones most likely to attract classification and evaluation duties, while documentation alone is unlikely to. So the two capabilities may end up on different sides of a line that does not yet exist.

Ask what the vendor's plan is for each pending framework, whether the copilot and the scribe are separable in contract and deployment, and whether terms allocate the cost of compliance changes that arrive after signature.

No fairness statement, subgroup analysis, accent or dialect performance disclosure, or evaluation methodology was located.

The absence sits directly underneath this vendor's founding thesis, which is what makes it the sharpest instance of its kind outside the largest vendors in this index.

The company's central argument is that clinical AI built elsewhere serves its region poorly, and that a system trained on regional literature, regional guidelines and local institutional protocols performs better for the populations it serves. That is a claim about differential performance across populations. It is precisely the proposition this axis exists to test, and the vendor has made it the basis of its business without publishing evidence for it.

The linguistic terrain makes it harder rather than easier. The three markets served span Brazilian Portuguese and several national varieties of Spanish, with substantial regional and social variation within each, indigenous language speakers, and clinical vocabulary that differs between countries. A single accuracy figure across that range would conceal more than it revealed, and no figure of any kind was located.

The evidence is also unusually obtainable here. Physicians at a named institution are reported using the product for around eight hours a day, which generates the volume that makes performance measurement across speaker populations feasible rather than anecdotal.

Ask what evaluation has been performed across languages, national varieties and regional accents, what the reference standard was, and whether the regional training claim has been tested against a system trained elsewhere.

The band is reached through the regime rather than through anything the vendor commits to, consistent with how this axis treats the non United States cohort. Brazilian data protection law treats health information as sensitive personal data, requires a defined legal basis for each processing purpose, and gives the individual enforceable rights over their own records including correction of inaccurate or incomplete data, with a supervisory authority behind them.

Comparable correction rights exist across the other markets served, though a buyer should confirm which regime governs its own contract rather than assuming uniformity. So the recorded patient has a mechanism that does not depend on the vendor offering one, and the vendor has a set of questions with concrete legal answers it should be able to supply. On the vendor's own side there is little, and one distinctive exposure that a buyer should size deliberately.

The system is grounded in each institution's own protocols, which is a genuine differentiator. It also means the system's recommendations are only as current as that ingested corpus, and a model still carrying a protocol the institution has since revised will keep recommending the superseded version confidently, with nothing published about how the corpus is refreshed or who validates that it matched the institution's intent in the first place.

That is a failure mode with no described control and it would be invisible to the clinician receiving the output. No accuracy figure, warranty, indemnity or remediation commitment was located, and the published assurance of the highest standards of clinical data protection names no standard. Ask who owns protocol currency, on what cycle, and what happens when a local protocol conflicts with a published guideline.

The vendor operates a dedicated interoperability product alongside the scribe and states that more than fifty systems are integrated, with information flowing automatically rather than being copied by hand.

That is a substantial claim and the count is higher than most vendors in this category can offer. It is not gradeable as delivered capability without the mechanism, and the mechanism is precisely what is unstated. No interface standard is named, no interoperability specification is referenced, and no distinction is drawn between systems reached through a permissioned connection and systems reached by driving a user interface. This index has recorded six vendors in this category that reach the chart the second way while describing it in language that sounds like the first.

The regional context makes the question sharper rather than pedantic. The markets served have less standardisation of health record interfaces than the United States or United Kingdom, where certification programmes and mandated interfaces give a vendor something concrete to build against. Achieving fifty integrations in that environment is genuinely difficult and is more likely to involve bespoke work per site, which raises different questions: what a given hospital's integration actually consists of, who maintains it, and what breaks when the underlying system is upgraded.

A buyer should therefore not read the count as a product specification. Ask what integration exists for your specific system, by what mechanism, whether it is bidirectional, what identity its writes carry in the audit log, and who is responsible for maintaining it.

No hosting region, residency option or subprocessor detail was located, and nothing establishes whether an external model service processes encounters.

Residency is a specific question here rather than a generic one, and it has a different shape from the single jurisdiction cases elsewhere in this index. The platform operates across three countries whose data protection regimes each impose their own conditions on international transfer of health information. A vendor headquartered in one of them, serving customers in the other two, is making cross border transfers as a matter of ordinary operation. So the question is not only where data sits but whether a Colombian or Mexican hospital's information is processed in its own country, in Brazil, or elsewhere entirely, and on what legal basis that transfer occurs.

That is answerable and consequential. A hospital's own regulator will ask it, and the answer determines whether the deployment is lawful rather than merely well engineered.

The stated expansion plan toward other emerging markets makes it worth establishing whether residency commitments are made per market or whether a single region serves everyone.

The interoperability layer adds a second dimension. Where the platform moves information between hospital systems automatically, a buyer should establish what transits the vendor's infrastructure in the process and what persists there afterwards, as distinct from what is merely passed through.

Ask for the processing location per country, the legal basis for any transfer between them, the subprocessor list, and which foundation model provider handles encounters.

Better than most in this index because tier structure is public even though enterprise pricing is not: a free tier exists for independent private practice physicians, with hospital system agreements negotiated separately. Publishing a free entry point tells an individual clinician exactly what they can access without a sales conversation, which is the same practice that earned Freed a high grade. Held back from A because no institutional pricing is disclosed.

Deliberately scoped to Latin America and localized rather than translated: models are trained on local institutional protocols alongside clinical guidelines and peer reviewed literature, which is what makes protocol adherence measurement possible. Deployed across more than 25 hospital systems in Brazil, Colombia, and Mexico.

A free tier for independent private practice physicians extends reach to clinicians outside institutional settings, which matters in a region the company cites as having 43 percent fewer physicians per capita than the OECD average.