AI has changed the economics of software creation. Code that used to take a quarter can be stood up in an afternoon. Unique workflows that were never worth building because the return did not justify the engineering cost are suddenly in reach. Ideas turn into working prototypes in hours.
And yet, inside the enterprise, the return on all of this has been disappointing. The industry data is blunt about it. Around 79% of organizations call AI adoption challenging despite significant investment, and only about 29% report substantial ROI. The teams that do see returns share one trait: they redesign the work rather than bolt AI onto it. Organizations that redesign the workflow are roughly three times more likely to see real ROI, yet only a minority actually do it.
The investment is not the problem. The missing redesign is. And there is a second, quieter reason the returns do not show up, one that is very familiar to anyone who has shipped enterprise software: a prototype is not a production system.
The hard part was never the code
A demo and a production system look similar in a screenshot and are worlds apart underneath. The distance between them is not the feature logic. It is everything the demo gets to skip.
In a regulated, people-sensitive domain like talent, the production system has to respect every user’s existing permissions, so a recruiter only has access to authorized information. Actions should be tracked in audit trails. Personal data has to be handled in a compliant manner.
The software has to stay up-to-date as the org changes, as a workflow changes, and as the model underneath it is updated. And it’s use is limited unless it is grounded in the actual intelligence of the domain: the skills, the roles, the matching logic, the career history that make a talent decision a good one rather than a plausible one.
Writing the code was never the bottleneck. The bottleneck is producing software that is governed, compliant-enabled, secure, domain-intelligent, and maintainable, and keeping it that way in production. That is the work AI did not eliminate. That is the work TalentForge is built to absorb.
We did not build a code generator. We built a foundry where business intent compiles into software that inherits the platform’s enterprise trust and talent intelligence.
How TalentForge works
TalentForge separates cleanly into two layers, and the separation is the whole idea.
The Talent Intelligence Platform. This is the layer Eightfold hosts, operates, and certifies. It holds the skills graph, job matching, career-path AI, talent profiles, learning, and the foundational talent data, and it exposes this intelligence for TalentForge through APIs and MCP tools. The Talent Intelligence Platform is multi-tenant, certified and audited, and where the network learning lives. With TalentForge, customers do not rebuild this and do not maintain it. They build on it.
The application layer. These are the apps and agents built on top of the platform foundation with TalentForge. You deploy them where you want, in your own infrastructure or on ours, and you control the design, experience and the data flowing through it. You own what you build. Eightfold owns and operates the intelligence underneath it. That boundary of Talent Intelligence Platform is what lets everything above it inherit properties that teams normally have to build from scratch.
Governed by default. Apps built on TalentForge automatically respect a person’s existing permissions through user-scoped sign-in. The permission framework is inherited. Actions are audited. Personal data is protected. The platform carries real certifications: FedRAMP, ISO 42001, ISO 27001 and SOC 2 Type II. In a hand-rolled build, this saves engineering time and risk. With TalentForge governance inherited from the platform is a default, not a project.
Talent-intelligent by default. The apps you build are not generic. They are built leveraging the talent intelligence platform’s domain expertise. A workflow built on TalentForge directly composes matching, the skills graph, and career trajectories. It is aware of the entities in your existing Eightfold implementation, so the code is grounded in your talent model rather than reimplementing a shallow copy of it. Building a hiring-manager agent means wiring real matching, sourcing, interview, and offer capabilities into the surface where the manager already works, not stubbing them.
Extensible beyond what Eightfold ships. TalentForge is not limited to the workflows we designed. Teams can provision entirely new entities, employees, org structures, and build entirely new applications on the same governed foundation. A leave-request entity can power a leave-management app. New people records power the complete HRIS, with permissions, governance, and compliance inherited rather than rebuilt. The foundation stays the same. The surface area of what you can build on the Eightfold platform does not have a fixed edge.
Consistent and maintainable. Builds follow the same design and architecture patterns, enforced through embedded skill files, so the output is uniform and reviewable rather than a pile of one-off prototypes with 50 different shapes. Build changes are versioned and iteration is safe, so you can evolve an app without breaking what is already live. This is what separates production-ready systems from prototypes.
What does TalentForge has to offer
TalentForge has two offerings: Foundry and Studio. TalentForge Foundry empowers our customers with tools for consistent, governed and compliance-enabled production ready software.
Headless Eightfold: With Foundry, we expose our Talent Intelligence Platform headlessly. We have built extensive APIs and MCP tools that can be consumed by apps and agents built with TalentForge. We also expose native Eightfold components so your app doesn’t look like an alien.
Forge Builder: You can use Forge Builder in Foundry to create and build custom apps and agents. After your app or agent is validated and approved by your admin, Forge Builder lets you deploy it within our secure cloud, or you can host it yourself. Built using native Eightfold platform components, your custom app fits perfectly within our current offerings. If you want your app or agent to follow your custom design and host it in your internal application, Foundry also supports that. At each step, you are in control: you decide what to build, how to build it, where to host it, and where to embed, and code ownership of what you build lies with you.
From intent to production, in one flow: The Forge Builder is an opinionated harness that is capable of turning a description into a talent‑intelligent app via a single guided lifecycle: ideation, requirements, design, build, and deploy.
Unlike traditional vibe coding that relies on conversation history, Forge Builder generates crystallized requirements (user stories) that carry through the process, from design to test cases to acceptance criteria, for production‑grade software. It includes built‑in rules to enable compliance, for inherited security, and proper governance.
Forge Builder aims to keep the app aligned with your defined requirements by following an iterative Build‑Test‑Fix cycle. Both builders and administrators can review the requirements, test cases, and security runs, which increase confidence in the resulting application. You also have full access to your app’s code and can perform code reviews.
It is LLM‑native. TalentForge runs on Anthropic’s Claude by default, supports other models, and lets you bring your own license.
TalentForge Studio: For teams that want to move forward without staffing the build themselves, the Studio offering provides more than software. We pair three roles on a build:
– An HR domain expert who translates the vision into a real workflow,
– A forward‑deployed engineer who codes it, sets up deployment, and integrates it with your systems,
– A compliance and AI‑governance expert who aligns the build with our Responsible AI framework.
With TalentForge, teams can build apps themselves using the APIs and Forge Builder, or engage Eightfold to build and operate it according to a roadmap. At the core is the same platform, intelligence, and trust; you decide how much you build versus how much you configure.
The software gets smarter as you build
Here is the property a general-purpose software foundry cannot offer, and it is the one that compounds.
Applications built on TalentForge feed data back into the intelligence graph. The skills graph gets richer, the matching gets sharper, the signals get deeper, and other apps on the foundation benefit from it. A performance workflow that captures conversational signals makes the talent picture better for a career app that never touched that data. Intelligence compounds across the network.
A foundry that produces isolated applications gives you N disconnected systems. TalentForge produces N applications that each make the foundation smarter and each inherit the accumulated intelligence of the others. The apps are the surface. The compounding foundation is the moat.
What teams are building
This is not a roadmap. It is in production.
Deutsche Telekom built a continuous performance and development experience shaped to their own workflows. In the words of Ilja Bitterling, their VP of Skills Intelligence and Performance Management: “Instead of forcing our processes to adapt to software, TalentForge gives us the opportunity to build solutions that fit our business, connect our workflows, and deliver experiences end to end.”
Deloitte built a branded recruiting experience with AI match scoring, compensation benchmarking, and market intelligence.
Eightfold runs a complete internal HRIS, with PII governance, built on the same foundation.
Across teams, the TalentForge Foundry is used to produce a hiring-manager agent living inside Teams, a talent agent that answers recruiters directly in Slack, a leave-management app, onboarding journeys with governed task routing, a real-time talent review with nine-box calibration, and continuous-development apps embedded in Career Hub. Three very different workflows solving three very different problems, with one governed, intelligent platform powering each.
We do not implement AI, we redesign the work
The constraints in enterprise software have shifted. It is no longer enough to ask whether a tool has a feature. The questions now are whether it scales and delivers the outcome, whether it is built on a dynamic AI-native foundation rather than a static system, and whether security, governance, explainability, and auditability are empowered like never before. A prototype fails the first. Bolt-on AI fails the second. Most hand-rolled builds fail the third.
TalentForge is our answer to all three. It is a way to take the work you actually want, redesigned for how AI can do it now, and turn it into production-grade, compliance-enabled, secure software, built on a talent intelligence foundation that is already trusted and that gets smarter as you build. From business intent to production code, on a foundation that compounds.
The enterprise talent stack is being rebuilt. We think it should be rebuilt by the people who do the work, on intelligence they can trust, without giving up the governance and security that production demands. That is the foundry we are building. If that is the kind of problem you want to work on, we are hiring.