One Model

One Model collects data from HR, recruiting, survey and finance systems into one store with rules on who sees what. It then builds dashboards, forecasts and an AI assistant on that data. It sells all three editions on a quote. The attrition-risk dashboards need its customer success team and a data engineer. Some AI features send data to OpenAI in the US when switched on. One Model tells customers to get internal approval before they create results that name individuals.

By Ashish Manchanda Published 10 Nov 2026 Last updated 10 Nov 2026

One Model brings your HR data into one store and sells by quote

In one line

One Model gathers your HR data from many systems into one store. It adds dashboards, forecasts and an AI assistant, and you ask it for a quote.

Buy it if

You run several systems, such as an HR system, a recruiting tool and a survey tool. You want one set of figures across them. You are ready to work with One Model’s customer success team on what to build. You will treat any result about who may leave as input for careful review.

Do not buy it if

You want a published price before any sales call. You cannot give the customer success team and a data engineer time for the attrition-risk dashboards. You have not agreed with your legal team who may see results about named people.

Price

Quote only. One Model sells Essentials, Data Mesh and Enterprise, and each is sold on a quote.

It connects your systems first, then builds dashboards, forecasts and an AI assistant

One Model is a data platform first and a set of dashboards second. The work of connecting your systems comes before the first chart. The platform has three parts. Data Mesh brings data in, cleans it and organises it. Data Stories turns it into dashboards. One AI builds forecasts.

Data reaches One Model through connectors. It can take headcount from Workday or SAP SuccessFactors. Labour costs come from finance systems. Recruiting figures come from Greenhouse or iCIMS. Oracle HCM Cloud data is extracted on a schedule you choose. Culture Amp and Qualtrics supply survey data. For Qualtrics, One Model ingests surveys rather than running them. Data from other sources can arrive as a CSV file or a file upload. It can also arrive on One Model’s hosted SFTP server, a secure server for file transfers. The data sits in storage behind a firewall and is processed in Amazon Redshift. Each customer gets a separate database.

On top of that data One Model offers over 500 standard metrics across 14 HR subject areas. Users can create or change metrics in the interface. One Model builds forecasts for hiring, attrition, capacity and workforce cost. It also builds employee-level attrition risk models. The attrition dashboard template has a by-name view of higher-risk employees. It also has an employee-level analysis for one selected person. That analysis gives the person’s predicted risk and the factors that pushed it.

The One AI Assistant answers questions in plain language and builds visualisations. One Model generates its insights itself. A language model only ranks them, and a separate one writes the business recommendation. One AI’s forecasting models are trained only on your own organisation’s data.

You can also connect your own AI tool. One Model for Enterprise AI uses an MCP server to bring workforce data into tools such as Copilot and ChatGPT. MCP is a standard link between an AI tool and a data source. One Model’s MCP link is in an early-access phase. Once data leaves One Model, your agreement with your AI provider governs it. A language model writes the final output. That always carries a small risk of translation errors or hallucinations, which are made-up answers. One Model recommends that you cross-check critical figures.

AI Impact is a newer product, announced on 6 October 2026. It measures how an organisation uses AI. It reads activity from OpenAI, Anthropic, Microsoft, Gemini, xAI, Glean and custom systems. It adds workforce context such as manager, role, function and cost centre. One Model sells it alone or alongside its Workforce Intelligence product.

Access rules go down to the row, and hosting regions include the EU

HR data is personal, so the first question is who can see what. This section covers four things you can check on One Model. They are its access rules and whether the AI assistant follows them. They also include where the data is hosted and how data moves out.

Managers can see only their own team

One Model can limit a manager to the people who report to them. A contextual rule does this, and it updates as your organisation data refreshes. Rules cover tools and data separately. Data Access Roles limit data at column level and row level, and a rule can be static or dynamic. That matters in HR, because a figure that is harmless in total can be sensitive for one person. One person can hold several roles, and then the rules combine. If two roles cover the same metric and one gives column access, the person gets column access and drill-through. Drill-through means opening the person-level records behind a figure.

The AI assistant follows the same rules

The assistant can only query data a person’s role allows. An Answer is visible only to users who can view the dashboard tile behind it. The same Data Access Roles and row-level rules apply when you connect your own AI tool through MCP. An assistant that ignored these rules would undo them.

You choose the region where One Model hosts your data

It can host in a region such as the EU. The options are US East, Canada, EU (Ireland) or Australia, set at onboarding. The AI features that use OpenAI are the exception, because OpenAI processes prompts in the US. One Model holds ISO 27001:2022 and SOC 2 Type II certifications. It encrypts data at rest with AES256 at a minimum and data in transit with TLS1.2 at a minimum. It adheres to the data processor requirements of GDPR. It will never process data in a way you have not requested or set up.

Data can move out to your own tools

You can download a CSV or schedule an export. You can send files to a destination such as an SFTP server, Amazon S3, Azure or Snowflake. You can also embed dashboards in other applications. One Model can push its full updated data model to your own data warehouse on Snowflake, Redshift or Azure. Power BI and Tableau can use it. Data also moves both ways with Databricks. Analysts can then work in tools they already know.

Firms from about 1,800 to over 130,000 staff use it

Buyers ask whether a platform works at their size and what customers got from it. Treat every figure below as the customer’s report or One Model’s, and test it in the demo.

Its customers include firms of very different sizes

Squarespace has about 1,800 staff and Maximus has 65,000. JLL has over 130,000. Paramount’s size is put at about 18,000 and also at 20,000 or more. Other customers include The New York Times, Dropbox, Reddit, NetJets, GE Aerospace and the American Cancer Society. The same platform runs at all of these sizes.

Most reported gains are hours saved

JLL reports data processing now takes a few hours, not about two days. Medibank found and removed 52 duplicate HR reports. A superannuation organisation reports that each hour in One Model replaced up to seven hours of manual reporting. One Model reports that Pure Storage saved years of team time and several hundred thousand dollars. It did so by investing in One Model rather than finishing an in-house project. If your team builds reports by hand, test these against your own figures.

Reported people results

Maximus reports that first-year voluntary turnover fell 70%, saving 5,000 or more jobs and $10M. Tabcorp reports more than 30,000 hours of manager time saved and $2.5 million in savings in its first year. NetApp raised dashboard usage from 10,000 to 90,000 views. A payroll services customer used predictive analytics to find about 2,200 employees at risk of leaving. Agios uses One Model dashboards to assess diversity outcomes through the hiring lifecycle. These are the customers’ own reports, and they sit closest to an HR decision.

Customers join data from many systems

Squarespace connected Greenhouse, UltiPro and facilities data in 2017. A payroll services company joined 200 or more data sources and saved more than 8,000 internal build hours. Joining systems is the job, so these examples show the range.

Going live has taken a few months

Robinhood had One Model up and running within three or four months. Elastic reports that implementation took only a few months from signing to launch. Set-up time decides when HR sees its first figures.

A supplier’s finances matter if you will rely on it for years. Chris Butler, David Wilson and Matthew Wilton founded One Model in 2014. They had worked at Infohrm, which SuccessFactors and SAP acquired. On 7 August 2023 One Model announced a $41 million investment led by Riverwood Capital. It reports becoming cash-flow positive in the second half of 2025. It also reports that 95% or more of its revenue comes from long-term subscriptions.

Set-up work, cut-off results, data sent to OpenAI and results that name people

These limits fall into two groups. Three concern set-up. They are building the first dashboards, incomplete results, and gaps in connectors, accounts and audit logs. Four concern data about people. They are results that name individuals and the law on decisions about a person. They also include data sent to OpenAI and AI Impact reading prompts. Each one decides who on your side has to be involved.

Attrition-risk dashboards need a data engineer

Building them means working with One Model’s customer success team. A data engineer also adds code to your processing script, the script that prepares your data. Plan for both before you promise HR a date. Customers with training can expose new fields. A support request can also do it, typically in two to five days. To use the One AI Assistant you contact your Customer Success Lead to enable it. An administrator then sets it up. One week every quarter, One Model’s support team pauses routine data set-up activities and meetings.

Large results can come back incomplete

One Model displays a message when a result goes past a display limit. The limit is typically 1,000 data intersections. The message tells you the data limit was reached and the results are incomplete. One Model adds that the results displayed may be incomplete or unreliable. A cut-off result can mislead anyone who misses the message. Ask how your largest reports compare with that limit.

Connector, account and audit gaps

SuccessFactors learning management system data is not available through the API. You must request it from SAP professional services. One Model does not typically need sensitive fields such as social security numbers. Customers should restrict them in SuccessFactors so they never leave it. Population restrictions set in SuccessFactors are not recognised by the API. With single sign-on, any employee who can reach your One Model site and signs in successfully gets a user account. Check what a new account can see by default. One Model keeps the audit history for as long as your copy of the system runs. It does not provide searching or filtering of the audit logs inside the application.

Named results come with conditions

The attrition dashboard template includes a by-name view of higher-risk employees. A Promotion Readiness model has one too. One Model tells customers to create by-name and employee-level results only with internal approval. It advises that person-level explanations should be inputs to a discussion. They are not automated decisions or guarantees. Drivers do not prove causation, so a factor behind a prediction is not proof of why someone leaves. Forecasts are projections from historical patterns. They are sensitive to data coverage, seasonality and organisational change. These results affect the people named, which is why the conditions matter.

The law can limit decisions about a person

Following One Model’s advice does not settle the legal question. GDPR Article 22 gives people a right not to be subject to a decision based solely on automated processing. That includes profiling with legal or similarly significant effects. Before a decision about a person leans on a by-name result, ask your data protection adviser to check Article 22.

AI features send data to OpenAI when enabled

The One AI Assistant uses OpenAI models through One Model’s enterprise account. OpenAI processes prompts in the US for One AI features where the subscriber enables them. Data sent to OpenAI endpoints is stored for 30 days for abuse monitoring and not used to train OpenAI models. Analyze sends the data behind a chart or table to the model with the request. Summarize a Storyboard page is in beta. Where a tile is split by a dimension holding personal information, those values may be sent. Dimension values are the labels a figure is split by, such as team or location. If personal data must not reach OpenAI, One Model advises keeping values that identify employees out of them.

AI Impact reads staff prompts

It analyses prompts from several AI tools. It also uses Microsoft Graph data about employees. It is not an employee monitoring tool. It is designed to understand AI adoption, activity and effectiveness. Prompt content is required for classification. Every prompt is classified across 14 dimensions, plus nine at session level. Managers can see individual prompt-level information where company policy allows. You can turn prompt-level reporting on or off at company or user level. You can also exclude employees in certain locations. A works council in Germany has a co-determination right over technical devices designed to monitor employees’ behaviour or performance. If you have one, ask whether AI Impact falls under it. Check who may see prompt-level information before you switch it on.

Three editions sold by quote, and what the contract covers

One Model sells three editions, Essentials, Data Mesh and Enterprise. Each is sold on a quote, so you learn the price by asking for one. What each edition holds matters more than its name, because forecasts and the AI assistant sit in different editions.

Essentials is pre-built insight packs. Extra data sources or insight packs are pay as you go. Data Mesh is for bringing data together across sources and governing it. Enterprise is end-to-end people analytics with AI. Dashboards come with Essentials and Enterprise. One AI predictive modelling comes only with Enterprise. The One AI Assistant is an add-on for Essentials and included in Enterprise. The MCP link to your own AI tool is an add-on for Essentials and Data Mesh and included in Enterprise.

One Model charges no per-query fee for MCP access. Use is subject to your contract. Your own AI tool needs enterprise licensing. Ask whether the Data Mesh edition comes with dashboards.

Only the Data Processing Agreement between you and One Model governs how it processes your data. Its sub-processors must delete or return all personal data when their engagement ends. They must certify deletion on request. Your own exit terms are a question for the order form. If you change HR system, One Model can store your historical data. It can also bring in data from the new software platform.

You could also get these reports and answers from agents built on your own systems. We have mapped what that looks like. The HR Agent Map

Ten questions for the demo

These questions fill gaps that a demo tends to skip. Each comes with the answer you want to hear. A vague reply is a reason to press.

Which edition holds the forecasts, the AI assistant and the link to our own AI tool? What does each extra cost? Good answer: A written list by edition, with the price of every add-on.
Does the Data Mesh edition come with dashboards? If not, what do they cost? Good answer: A written yes or no, with the price of the dashboards if they are extra.
Who builds our attrition-risk dashboards, whose data engineer adds the code, and how many days does it take? Good answer: A named advisor, a named engineer and a go-live date in the order form.
What does a limit of 1,000 data intersections mean for our biggest reports? Good answer: A worked example from our own data, and the reports that would hit it.
Can we switch off every by-name and employee-level view, and who can switch them back on? Good answer: Off from the start, one named administrator can change it, and the audit history records it.
When someone signs in with single sign-on, what can they see before an administrator gives them a role? Good answer: Nothing, until a role is assigned.
How do we search the audit history? Good answer: A full export on request, in a stated format.
Which features send data to OpenAI, and can we turn each one off? Good answer: A named list of features, each with its own switch, off until we ask.
For AI Impact, which AI providers do we connect and how much detail do we get from each? Good answer: A named list by provider, with the level of detail written down.
How long is the agreement, and what do we get back when it ends? Good answer: A stated term and a full export in a named format by a set date.

Alternatives to One Model

If One Model leaves your shortlist, Visier and Crunchr also join data from several systems. Visier’s connectors join HR, payroll, recruiting, talent and data warehouse data into one layer. Workday, SAP and Oracle sell analytics built around their own HR systems. Orgvue looks at how an organisation is shaped. Using what you already own is the last option.

Visier and Crunchr raise a question about how the price grows. Visier prices by the number of employees and by the products bought. Crunchr charges one fee for the full platform, based on the size of the workforce, with unlimited users. Both tell you to contact them for a price. Visier also offers a 30-day free trial with sample data or your own. Ask One Model whether its quote follows headcount.

Crunchr and One Model both keep your history through a change of system. Crunchr does it once the new system is mapped to its data model.

Workday People Analytics analyses Workday HCM and talent data. You choose 4 to 36 months of history to load, and it analyses up to 400 thousand employees. SAP People Intelligence is an add-on licence to SAP Business Data Cloud. It connects to SuccessFactors data and beyond. NTT DATA Business Solutions is a SAP reseller. Its UK G-Cloud list price is £2.30 per user per month for 1 to 2,000 users. The price falls with volume and rises 3.3% a year. Oracle Fusion HCM Analytics is a prebuilt solution for Oracle Fusion Cloud HCM. Oracle offers to get you started with it for free. Orgvue covers structure, spans, layers and costs.

If you already run SAP SuccessFactors, Story Reports are available as part of the platform at no additional cost. Embedded reports in Employee Central are included at no additional cost too. Your own warehouse is another route if you already hold the data there. Our guide for teams with their own AI and data warehouse covers it. The people analytics guide and the comparison page set One Model beside the rest.

More in this guide

The People analytics guide
Compare the 33 researched products what each brings together, what its AI does, what it costs and the contract terms.
Mid-market companies which people analytics products suit 200 to 2,000 people, what they cost and what to ask.
Enterprise companies which people analytics products suit 2,000 people or more, what they cost, how contracts run and what to ask.
Companies whose HR system has analytics what the analytics in your HR system covers, what it costs and when to buy something separate.
Companies using their own AI tool or data warehouse how each product lets your own AI tool or data warehouse use the data, what it costs and what to check.