Workheld AI
Workheld AI is the AI agent inside the Workheld industrial platform from Austrian publisher Workheld GmbH. It gives field technicians instant chat answers on maintenance and production, predicts failures, and feeds existing ERP, MES and CMMS systems.
What is Workheld AI?
Workheld AI is the artificial intelligence component of the Workheld platform, published by Workheld GmbH in Vienna, Austria. The platform has three parts: Workheld Flow for digital workflows and mobile field work, Workheld Sense for industrial IoT, and Workheld AI, the agent itself. A fourth module, DeepConnect, opens secure remote access to edge equipment over HTTP, VNC and SSH from Sense. The publisher positions the whole as Connected Workforce Management for Industry 4.0, at home in Europe, globally available. The agent's promise is stated plainly on the product page: Workheld AI makes knowledge available where it is needed, giving technicians and service teams instant answers directly via chat in the Workheld app, with no detour through coordinators. It addresses three problems the publisher names, namely process complexity and data silos, gaps in knowledge transfer between teams and generations, and the time and cost of breakdowns and maintenance. Four capabilities are announced. As an intelligent assistant, the agent answers technicians immediately and guides them step by step. As knowledge management, it opens up experience knowledge and documentation. As predictive maintenance, it detects failures early through automated analysis, spotting patterns and optimisation potential in maintenance data. And as an integration layer, it connects to existing ERP, MES and CMMS systems. A Best Match function automatically assigns work orders to the best-suited employee, simplifying team planning. The product page names four audiences: production, for efficient shift handovers; maintenance, for fast fault diagnosis and standardised procedures; asset management, for transparent data behind investment decisions; and mechanical engineering, for connecting people, machines and data. Technically, model calls run through Microsoft Azure OpenAI Service inside Workheld's own Azure tenant in the West Europe region, whose data centres are in the Netherlands, with no transfer to OpenAI Inc. The platform can also run in the customer's own Azure tenant, on-premises on company servers, or on SAP HANA and SAP Cloud Platform. Flow itself runs on the web, as a desktop application and as iOS and Android apps, with an offline mode for writing reports without a connection. Workheld GmbH claims up to 30% more productivity and up to 25% lower costs. The commercial entry point is a free AI Readiness Check.
What it does
- Ask a technical question in chat and get an immediate answer inside the Workheld app
- Follow step-by-step instructions while carrying out a job on site
- Retrieve experience knowledge and documentation without going through a coordinator
- Receive real-time AI suggestions for fault diagnosis and reporting
- Detect failures early through AI analysis of maintenance data (predictive maintenance)
- Assign a work order automatically to the best-suited employee with the Best Match function
- Feed the resulting data into an existing ERP, MES or CMMS system
When to use Workheld AI / When not to
A quick filter to help you decide if Workheld AI is the right fit.
When to use Workheld AI
- Industrial companies whose technicians work in the field, in after-sales service, installation or commissioning, and who need answers on site without calling a coordinator.
- Maintenance organisations looking for faster fault diagnosis, standardised procedures and early detection of failures through AI analysis of maintenance data.
- Teams already running or evaluating Workheld Flow, since the agent lives inside the Workheld app rather than as a separate product.
- Machine builders and industrial asset managers who want people, machines and data connected, and transparent figures behind investment decisions.
- European organisations with strict GDPR requirements, served by Azure West Europe hosting, ISO/IEC 27001 certification and a published processing agreement, including customers who need their own Azure tenant or an on-premises install.
When not to use Workheld AI
- Individuals and small teams who want to sign up online: there is no self-service registration, only a demo form or an AI Readiness Check.
- Buyers who need a published price, a rate card or a free plan before entering a sales conversation.
- Anyone looking for a standalone AI assistant: Workheld AI is a component of the Workheld platform alongside Flow and Sense, and is not sold on its own.
- Developers expecting public API documentation, which the website does not publish even though the vulnerability disclosure policy lists public API endpoints in its scope.
- Use cases involving special categories of personal data under Art. 9 GDPR, or prohibited (Art. 5) and high-risk (Annex III) AI Act scenarios, which the processing agreement excludes unless expressly agreed.
How to use Workheld AI
A typical end-to-end flow, from setup to results.
- Book a free AI Readiness Check or request an individual demonstration through the website form, as there is no self-service sign-up
- Use that first meeting to go through your company and the AI potential of the business area concerned
- Agree an individual offer with Workheld, since no price list is published
- Choose a deployment: Workheld's cloud on Azure West Europe by default, your own Azure tenant, on-premises servers, or SAP HANA
- Have your company administrator create the user accounts (first name, last name, email, optional phone number, company, role)
- Log in at flow.workheld.com, with authentication through Auth0/Okta and optional single sign-on via Microsoft Entra ID
- Open the Workheld app and work with the agent through its chat interface
- Have the import and export interfaces to your ERP or CRM built, their scope depending on the system and the requirement profile, with Workheld advising on the development
- For Sense and Luna, supply the edge hardware, the IT/OT network segmentation and the firewall rules for the outgoing MQTT connection
- Raise questions and incidents through Workheld's Atlassian portal, which combines a service desk and a knowledge base
Pros & Cons
Pros
- AI processing stays inside the publisher's own Microsoft Azure tenant in the West Europe region, with no transfer to OpenAI Inc.
- A written and repeated commitment that inputs are not used to train the underlying language models
- ISO/IEC 27001:2022 certified by an accredited body, with the certificate and the ISMS policy downloadable as PDFs
- The Art. 28 processing agreement is published in full, with detailed technical and organisational measures and a named subprocessor list that includes external development partners and their countries
- Intellectual property in both AI inputs and outputs stays with the customer, data is returned in JSON or CSV within 30 days, the first export of each contract year is free, and breaches are notified within 24 hours
- AI Act transparency, with an Art. 50(1) notice and an explicit split of provider and deployer roles, alongside a vulnerability disclosure policy with safe harbour, CVSS-based SLAs and a dedicated address
- A choice of hosting (Workheld's cloud, the customer's Azure tenant, on-premises or SAP HANA) behind a mature field platform with iOS, Android and desktop apps and an offline mode, from a publisher active since 2015 with named industrial references
Cons
- No public pricing at all: no pricing page, no rate card, not even an entry-level figure
- No free plan and no self-service trial, so access always goes through a form and a sales conversation
- The AI agent cannot be bought on its own, being a component of the Workheld platform
- No public API documentation, even though the vulnerability disclosure policy lists public API endpoints in its scope
- The website is unevenly bilingual: the reference content is German, and the German version of the processing agreement legally prevails
- The AI product page stays short and general, with nothing on the models used, the context window or volume limits, and no interface language is declared anywhere
- Support has no email address and runs only through the Atlassian portal and the contact form, the contractual liability cap is low, an external development partner in Ukraine can access EU-hosted personal data remotely, and there is no paid bug bounty
Pricing & Plans
There is no free plan and no public price. Workheld publishes no pricing page: none of the 62 URLs in its sitemap covers pricing, and the addresses /en/pricing/, /preise/ and /en/preise/ all return 404. The official FAQ states that various packages exist depending on factors such as the size of the company, and that Workheld makes an individual offer on request. The general terms and conditions describe the possible billing models without setting any amount: a fixed recurring charge per registered user and month, as a flat rate regardless of the number of transactions; an extension of that model in which the amount depends on the functional scope actually used; or other models based on data volume, transactions, consumed CPU hours, or a constant price over a contract term. Prices and payment terms are fixed in the offer, implementation costs may be invoiced separately, all prices are stated in euros excluding VAT, and recurring fees are indexed to the consumer price index published by Statistik Austria. Use of the platform requires a valid licence or a test phase that has been expressly agreed.
- there is no Starter
- Pro or Enterprise tier anywhere on the website
- a fixed recurring amount for each registered user
- as an unlimited flat rate regardless of the number of transactions (GTC 1.6.1)
- an extension of the per-user model in which the amount depends on the functions actually used (GTC 1.6.2)
- charging by data volume
- by transactions
- by consumed CPU hours
- or a constant price over a contract term (GTC 1.6.3)
- Workheld's cloud on Azure West Europe
- the customer's own Azure tenant
- on-premises servers
- or SAP HANA
- implementation costs
- consulting
- interface development and a two-day Digital Maintenance Check Workshop
- Flow
- Sense and AI are contractually distinguished as components of the same platform rather than sold as separate products
Data, GDPR & hosting
A consolidated view of how Workheld AI handles your data.
GDPR overview
GDPR implementation is documented in unusual detail. Workheld GmbH is established in Austria, so no Art. 27 representative is required, and it has appointed a named data protection officer, Dipl.-Inf. Christine Geier, MBA, reachable at hallo@workheld.com or by post in Vienna. The privacy policy lists the rights of access, rectification, erasure, restriction, portability, objection, withdrawal of consent, freedom from automated decisions and complaint, with a one-month response time extendable by two months, and names the Austrian Datenschutzbehoerde as supervisory authority. A model Art. 28 processing agreement is published and downloadable, effective 13 May 2026, with technical and organisational measures, a named subprocessor list and breach notification within 24 hours. Processing stays in the EEA, and third-country transfers rely on an adequacy decision or standard contractual clauses. Workheld is certified ISO/IEC 27001:2022 (AT-IS-20260241) and publishes a vulnerability disclosure policy with safe harbour.
Who owns the data?
Under the published Art. 28 agreement the customer is the controller and Workheld only the processor. Clause 19b(iii) states that intellectual property rights in inputs and in AI-generated outputs remain with the controller, as far as legally permissible, and Workheld derives no usage rights from the AI feature beyond performing the contract. Clause 6 forbids using the data for its own or third-party purposes, or passing it on without written instruction, and the privacy policy adds that Workheld neither sells personal data nor discloses it for advertising or marketing. At the end of the contract, data is returned or destroyed within 30 days in a structured, machine-readable format. The customer's own administrator can view and manage the data its users generate.
Reuse rights
Platform data serves service delivery, authentication, rights management, billing, stable operation, abuse detection, pseudonymised product analytics and, on consent, marketing communication, on the legal bases of Art. 6(1)(b), (f), (c) and (a) GDPR, with purpose limitation explicitly stated. For the AI feature, inputs are processed inside Workheld's own Microsoft Azure tenant in the West Europe region and model calls run through Azure OpenAI Service. The publisher states that inputs are not used to train the underlying language models and that nothing is transferred to OpenAI Inc. An AI Act Art. 50(1) notice warns that the user is interacting with an AI system and that generated results may contain errors and do not replace expert review by authorised personnel. Product analytics run on pseudonymous data in Mixpanel, Sentry is configured with sendDefaultPii set to false, and Session Replay masks entries and personal content, covering 10% of sessions and 100% of error cases. The customer keeps its data and the AI outputs and can reuse them without asking permission, exporting them in JSON or CSV, with the first export of each contract year free of charge.
Data retention & training
Hosting summary
The Workheld platform runs on Microsoft Azure in the West Europe region, with the underlying data centres located in the Netherlands; the publisher's support FAQ names a data centre in Amsterdam. The contracting entity is Microsoft Ireland Operations Limited in Dublin. AI model calls go through Azure OpenAI Service in the same tenant and the same region. Backups are geo-redundant within Azure West Europe, with DDoS protection, a web application firewall, regular restore tests and TLS 1.2 or higher in transit. The HiveMQ MQTT broker and the TimescaleDB time-series database are self-hosted by Workheld in that region, so their vendors have no access. Processing takes place in the EEA as a matter of principle, and third-country transfers occur only under an adequacy decision or standard contractual clauses. Several subprocessors host in the EU but have US parent companies, among them Auth0/Okta, Sentry, HubSpot, Mixpanel and SendGrid/Twilio Ireland, while 84codes AB (CloudAMQP) is Swedish. Customers may instead deploy in their own Azure tenant, on-premises, or on SAP HANA. The marketing website is hosted separately, on WordPress at InstaWP in an EU data centre, resolving to an address operated by Hetzner in Nuremberg, Germany.
Things to keep in mind
Risks and trade-offs to weigh before adopting Workheld AI.
- No public price means nothing can be budgeted before a sales conversation, and recurring fees are indexed to the Austrian consumer price index
- The publisher warns that AI-generated results may contain errors and do not replace expert review by authorised personnel; treating a chat answer as a verified work instruction on a machine is a real safety risk, and leaning on the assistant can erode the very diagnostic skill it is meant to preserve
- The signed processing agreement prevails in German, the English version being a convenience that gives way in case of conflict
- A development subprocessor established in Ukraine, BigTree108 in Ivano-Frankivsk, can remotely access personal data hosted in the EU, under standard contractual clauses and additional measures
- The contractual liability cap is low, at 10% of the contract value per claim and a maximum of 15,000 euros
- The customer is designated deployer under the AI Act and carries the Art. 50(1) transparency obligation itself, and entering special categories of personal data under Art. 9 GDPR into the AI is contractually forbidden without express agreement
- The processing agreement page is reachable from no navigation link at all, only through its own URL, and the website's own host in Nuremberg must not be confused with the platform's hosting on Azure West Europe
Setup & Integrations
Technical difficulty
Moderate to demanding, and never self-service. Onboarding starts with a demo form or an AI Readiness Check, then a commercial offer, and accounts are created by the customer's administrator. Authentication runs on Auth0/Okta with optional single sign-on via Microsoft Entra ID. Connecting an ERP or CRM requires building import and export interfaces whose scope varies by system, and the terms put test, integration and production environments, API access, service accounts, VPN links and identity-provider configuration on the customer. Sense and Luna add edge hardware, IT/OT network segmentation and firewall rules. A two-day Digital Maintenance Check Workshop is offered beforehand.
Deployment
Apps stores
Integrations
Behind Workheld AI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is Workheld AI?
How much does Workheld AI cost?
Is there a free trial?
Where is the data hosted?
Are my inputs used to train the language models?
Who owns the data and the AI-generated outputs?
Is a data processing agreement available, and are subprocessors disclosed?
Is Workheld certified?
Can Workheld AI be connected to an existing ERP, MES or CMMS?
Is there a public API, and how do I get support?
Should you pick Workheld AI?
Workheld AI is convincing on data governance and frustrating on price transparency. On governance, almost everything a European industrial buyer normally has to ask for is already published. The platform runs on Microsoft Azure in the West Europe region with data centres in the Netherlands, model calls go through Azure OpenAI Service inside Workheld's own tenant, inputs are explicitly not used to train the underlying models, and nothing is transferred to OpenAI Inc. A full Art. 28 processing agreement is downloadable, with detailed technical and organisational measures and a named subprocessor list. Intellectual property in inputs and AI outputs stays with the customer, and the publisher holds an ISO/IEC 27001:2022 certificate valid to February 2029. On price, the opposite is true. There is no pricing page, no rate card, no free plan and no self-service trial. Everything happens after a form, an AI Readiness Check and a commercial conversation, and the terms merely list the billing models that could apply. Budgeting in advance is impossible. Two further points deserve attention. The agent is not a standalone product: it is a component of the Workheld platform, next to Flow and Sense, so adopting it means adopting the platform. And the AI product page itself stays general, with nothing on the models, the context window or volume limits, which leaves far less to verify than the legal documentation offers. For a European industrial company already digitalising its field interventions, and especially one with strict data protection requirements, Workheld AI is a serious candidate, since the field platform underneath it is mature, with iOS, Android and desktop applications and an offline mode. For anyone who wants to weigh a price, a trial or the technical characteristics of the AI before talking to sales, there is nothing yet to work with.
- Choosing a selection results in a full page refresh.
- Opens in a new window.