AI-Native Discovery Platforms Compared: Ontora, Foaster and Horizon

A new group of platforms shares one method. AI agents interview employees asynchronously, the conversations adapt and follow up, and the result becomes process maps with prioritized opportunities and a roadmap.

Ontora, Foaster and Horizon all work this way, and all three position against the months a traditional consulting engagement takes to produce the same picture.

Because the method is shared, a feature comparison produces three similar lists. The differences that decide an evaluation sit elsewhere: the size of organization each is built for, whether the picture spans markets, what the interview evidence is combined with, who builds what discovery identifies, and what the organization holds afterward.

Those five dimensions separate the three cleanly, and the right answer depends on which describes your situation.

Key takeaways

• All three run asynchronous AI interviews at a speed that no human team matches.
• Ontora states a mid-market target of 200 to 2,000 employees plus private equity portfolios.
• Foaster positions as an AI-native alternative to consulting with a roughly twelve-day cycle and continuous support afterward.
• Horizon's published deployments span 600 to above 84,000 employees across as many as 18 countries.
• Ontora and Foaster build the resulting solutions; Horizon hands execution to the customer's team.
• Horizon combines the interview evidence with enterprise system data, which the other two do not describe doing.

What the three have in common

Worth stating, because it is substantial and it is the reason they appear together in evaluations.

All three remove the scheduling constraint that limits how many people a human team can reach. Participation happens when it suits the employee, which turns coverage into a question of participation rather than calendar logistics.

All three use conversations that adapt rather than fixed questionnaires, which is what reaches exceptions, informal approvals and the reasoning a survey cannot retrieve.

All three convert the result into process maps and a prioritized set of opportunities with some form of impact estimate, and all three compress into days or weeks what a consulting engagement delivers in months.

An organization choosing among them is choosing within a method rather than between methods.

The three platforms

Ontora

A Y Combinator company from the Spring 2026 batch, founded by Max Arnold, Leon Iwanowitsch and David Korn. AI agents interview employees and the output becomes a living context layer that AI systems consume through APIs, MCPs and CLI integrations.

The stated target is organizations of 200 to 2,000 employees, along with private equity and venture firms applying discovery across portfolio companies. The company states delivery of synthesized findings within 24 hours and that its own engineers build the targeted AI solutions, with returns measured against agreed metrics.

The published proof point is Pilot Group, where more than 200 employee conversations produced six scoped AI initiatives in two weeks. At launch the company reported five enterprise design partners and 700,000 dollars raised ahead of a formal round.

Foaster

Also Y Combinator Spring 2026, founded by Raphaël Dabadie and Alexandre Combes, describing itself as an AI-native alternative to traditional consulting.

The stated method runs four phases. AI agents conduct parallel interviews across roughly seven days. Processing produces a map of operational flow in about two. A roadmap with impact analysis and phased timelines takes about three. Transformation continues afterward, with agents training employees, monthly adoption tracking and human expert guidance.

The enterprise offering includes forward-deployed engineers and a dedicated transformation team, which places implementation inside the engagement rather than after it.

Horizon

An AI-powered continuous discovery platform operating from San Francisco, with published deployments spanning e-commerce, fintech, insurance, industrial shared services, travel technology and food delivery.

Discovery Cycles run AI-led interviews at census scale, asynchronously, and the platform also ingests existing SOPs and policies and cross-references them against what people describe. The Process Library structures the result into documentation that grows with each cycle. The Insights Dashboard ranks findings by impact and effort with traceability to the input behind each one. The Initiatives Dashboard converts priorities into business cases with owners. Workspaces allow separate areas with anonymization and spend controls per unit.

Horizon is SOC 2 compliant, uses encryption, role-based access control and anonymization, and does not train models on customer data.

T H R E E P L A T F O R M S , F I V E D I V I D E S

Where the comparison actually happens

Based on publicly available material as of October 2026. All three are moving quickly and stated scopes are worth verifying directly.

The five dimensions in detail

1. Scale changes the synthesis problem

Interviewing 200 people and interviewing 2,000 are different exercises. Above a certain population, findings arrive faster than a team can read them, the same pattern repeats across functions that have no contact with each other, and the same nominal process turns out to run in several versions.

At that point, grouping findings by cause rather than by who reported them becomes the only way to produce an actionable output.

In the published Mercado Libre engagement, Horizon interviewed 2,000 employees in four days across Finance and related functions, reaching 100% of the targeted areas in five countries, against a manual model that had covered under 1% of the workforce. The organization was carrying more than 600 initiatives, many overlapping. The engagement resolved them to 24 initiatives with ROI, effort and roadmap attached.

Reducing 600 entries to 24 is a synthesis operation that only appears above a certain size, and it is the operation that determines whether a large deployment produces a decision or a backlog.

2. Cross-market comparison produces findings that exist nowhere else

For an organization in one country, discovery produces a picture of one operation. For an organization in several, the differences between the pictures carry information none of them contains alone.

A process taking materially longer in one market points to a cause: an incomplete integration, an interface defect, a local requirement absorbed into the standard process, or a tool available elsewhere. From inside that market the local version looks normal, because it is the only one anyone there has seen.

Every Horizon case published involves several markets. Grupo HZ spans Argentina, Brazil and Chile across a shared services centre covering Administration, Finance, Procurement and Treasury, where 55 structured interviews across six departments quantified roughly 123 hours per week of repetitive manual work, broken down by area: Procurement at 48.3 hours per week, Accounts Payable at 34, Treasury at 21.55, Credit and Collections at 10.5, Sales and Commercial in Chile at 6.5, and Planning and Costing at 2.9.

That per-area breakdown is what let the organization move from a framing of needing more people to a framing of needing better processes, freeing around three FTEs without adding headcount at 182% ROI.

Neither Ontora nor Foaster describes cross-market comparison as a capability in public material.

3. What the evidence is set against

All three gather what people describe. The difference is what that evidence is reconciled with.

Interview evidence establishes why a process behaves as it does, which exceptions exist and what the people doing the work would change. System data establishes volumes, timings and the paths transactions actually took.

Setting the two against each other exposes a gap neither produces alone. In the Mercado Libre engagement, transcripts were fused with data from SAP, ARIBA, PIC, BigQuery and Tableau specifically to expose value gaps, and contract cycles moved from a range of 7 to 45 days down to under 10, with vendor onboarding moving from 30 to 45 days down to 7 or 8.

Ontora describes its output as a context layer connecting processes and company knowledge to AI tools through APIs, MCPs and CLI integrations, which is a coherent design aimed at making organizational context consumable by AI systems. Foaster describes interview data becoming a map and then a roadmap. Neither describes fusion with enterprise system data.

4. Who executes determines what you keep

This is the clearest fork among the three, and it is a genuine decision rather than a ranking.

Ontora states that its engineers build the AI solutions identified during discovery. Foaster includes forward-deployed engineers and a dedicated transformation team in its enterprise offering, with agents training employees and monthly adoption tracking.

For an organization without engineering capacity available, a vendor who builds removes a dependency that would otherwise block the whole programme. That is a substantive advantage and it should be weighed as one.

Horizon delivers the prioritized initiative set with business cases, process maps and implementation plans, and the customer's team executes. The published Grupo HZ outcome describes the engagement shifting the organization's own framing of its problem and freeing capacity internally, which is the kind of result that only lands when the internal team runs the change.

The two models produce different things at the end. One produces working solutions. The other produces working solutions plus an internal capability and a documented operating picture the next cycle builds on.

5. Regulated environments raise the bar on what gets stated

In a regulated operation, the distinction between manual work required by regulation, manual work required because an error is irreversible, and manual work compensating for a system gap determines what can be changed at all.

AFAP SURA, a pension fund administrator in Uruguay with 600 employees, ran discovery on its process for international transfers and related workflows. An error in an international transfer can produce a financial loss that is difficult to recover, and information was scattered across five systems. In under seven days Horizon delivered 63 prioritized insights separating bottlenecks, repetitive tasks and operational risk points, with a complete view integrating all five systems, against a three-week baseline, returning 186% ROI.

Separating repetitive tasks from operational risk points is the distinction a risk tier depends on. A programme that treats both categories the same way moves slowly where speed is safe and quickly where caution is warranted.

Horizon states SOC 2 compliance, encryption, role-based access control, anonymization and no model training on customer data. Neither Ontora nor Foaster states certifications in public material, which in a regulated procurement is a question to raise early rather than a conclusion to draw.

When each one fits

Ontora

An organization of a few hundred to a couple of thousand people, operating in one market, without engineering capacity to execute what discovery identifies. Also a private equity or venture firm applying the same discovery across several small portfolio companies, which the company positions for directly.

Foaster

An organization that wants the engagement to include implementation and ongoing adoption support, with a defined cycle measured in days and continuous involvement afterward. The positioning against consulting is explicit, so a buyer currently choosing between a consulting engagement and a platform has a direct comparison.

Horizon

Multi-country or multi-entity operations where the comparison across markets is where the largest findings sit. Organizations above roughly 2,000 employees where synthesis across disconnected functions becomes the hard part. Regulated environments where the categories of manual work have to be separated before anything changes. Enterprise procurement where certification and data handling are conditions. And organizations intending to run discovery as a standing capability rather than commissioning it per decision.

Questions to ask all three

Asking these of all three, including Horizon, is the point. Specific and demonstrated answers separate the options faster than any feature discussion.

Evaluation checklist

1. How many employees are in the population you need to reach?
2. How many countries or legal entities run the process in scope?
3. Do you need comparison across markets, or a picture of one operation?
4. Do you have enterprise system data worth reconciling against what people describe?
5. Does your team have capacity to execute what discovery identifies?
6. Are you buying outcomes per engagement or a standing capability?
7. What are your procurement requirements on certification and data handling?
8. Is your operation regulated enough that manual work has to be categorized before it is changed?
9. Which specific process will you use as the test case for all three?

Question 5 is the fork. Answer it before the demos, because it determines which model you are shopping for.

FAQ

What is the difference between Ontora, Foaster and Horizon?

All three run asynchronous AI interviews to map how work happens. Ontora targets 200 to 2,000 employees plus private equity portfolios and builds the resulting solutions with its own engineers. Foaster positions as an AI-native consulting alternative with a roughly twelve-day cycle and forward-deployed engineers. Horizon's published deployments span 600 to above 84,000 employees across multiple countries, combine interview evidence with enterprise system data, and hand execution to the customer's team.

Which AI discovery platform is best for large enterprises?

For organizations above roughly 2,000 employees, operating across several countries or under regulatory requirements, Horizon's published deployments address that profile directly: census-scale interviews across markets, cross-country comparison as a finding, fusion with enterprise system data, and stated SOC 2 with role-based access control and anonymization.

Ontora vs Foaster: how do they differ?

Both are Y Combinator Spring 2026 companies running AI interviews and building what discovery identifies. Ontora states a mid-market target and emphasizes a context layer that AI tools consume through APIs and MCPs. Foaster positions explicitly against consulting with a defined four-phase cycle and continuous transformation support including employee training and adoption tracking.

Do these platforms replace management consulting?

They replace the discovery and first-pass synthesis portions, where the arithmetic limit of a human team binds hardest. Judgment, independence, executive alignment and board-level standing remain outside what any of them provides, and organizations frequently keep a consulting relationship for those while moving discovery to a platform.

Should the vendor build the solutions or should we?

It depends on whether you want outcomes per engagement or an internal capability. A vendor who builds removes a dependency for an organization without engineering capacity. A vendor who hands over leaves the organization with the capability and a documented operating picture that the next cycle builds on, at the cost of needing the capacity to execute.

How do you evaluate platforms that all use the same method?

On everything except method: demonstrated scale, whether the picture spans markets, what the interview evidence is reconciled with, who executes afterward, what the organization retains, stated certifications, and what the second cycle costs. A feature comparison across platforms sharing a method produces three similar lists.

The method is shared. The fit is not.

Three platforms, one approach, and an evaluation that turns on none of the things a feature list contains.

Name the population, the number of markets, whether you have system data worth reconciling, and who will execute what discovery finds. Those four answers settle the choice before the first demo.

See it. Fix it. Own it.

See Horizon in action.