Translation Management Systems

The evaluation this category was waiting for has been out since September 2025, and it is the first scorecard the localization technology market has ever had. The Forrester Wave: Translation Management Systems, Q3 2025, scores twelve vendors on twenty six criteria, and its central finding renames what the market sells: "AI reshapes localization workflows, not just translations."

The inaugural edition: The Forrester Wave: Translation Management Systems, Q3 2025

Principal analyst Kathleen Pierce announced the first-ever evaluation on September 10, 2025, scoring twelve vendors against nineteen current offering criteria and seven strategy criteria: Bureau Works, Centific, LILT, Lionbridge, Lokalise, memoQ, Phrase, RWS, Smartling, TransPerfect, Unbabel, and XTM.

The field is the market's structure in one roster. The language service giants, Lionbridge, RWS, TransPerfect, sit alongside the software vendors, Phrase, memoQ, XTM, Lokalise, and the AI-first challengers, LILT, Unbabel, Smartling. TransPerfect's acquisition of Unbabel was announced just before publication, and the two were still scored separately, the market's consolidation happening in the field's own roster.

Pierce's framing defines what the category is for: TMS platforms are "vendor-neutral orchestration layers" that unify fragmented translation technologies, and integration is "the backbone of modern TMS platforms," connecting content management systems, digital asset management, development environments, AI providers, machine translation engines, and iPaaS.

The Leader with twenty one of twenty six

Phrase took a Leader placement with the highest possible scores in twenty one of twenty six criteria, including vision, innovation, workflow automation and AI agents, quality standards and estimation, integration breadth and depth for developers, scalability, and roadmap.

Forrester's evidence for Phrase runs on the numbers: strategy success showing in growth, retention, and partnerships, reference customers praising its "practical, AI-forward thinking," and the figure that defines the edition's promise, AI agents automating ninety nine percent of workflows for some reference customers.

The fit verdict: "medium-to-large enterprises with high localization maturity and complex workflows seeking fully automated, integrated, AI-powered localization." The verdict's honesty is the buyer's warning: the Leader's platform is built for organizations that have already industrialized localization, and the automation it sells presumes the maturity.

The other two Leaders and their different integrations

XTM took a Leader placement on its roadmap for orchestration and AI agents, workflow controls, financial management and reporting, linguistic asset management, and governance, positioned for regulated industries.

LILT took the third Leader placement, cited for uniting custom AI models with human-verified precision, scoring highly on workflow automation, AI agents, transcreation and adaptation, quality standards, compliance, and innovation.

Three Leaders, three theories of the same engine. Phrase sells the fully automated pipeline. XTM sells the governed, financially controlled pipeline. LILT sells the custom-model-plus-human pipeline. The scorecard's top rung holds all three because the market genuinely runs all three, and the difference between them is the buyer's actual decision.

The lock-in the report named

The report's sharpest negative finding is about switching. Many reference customers wanted to switch providers but were held back by integration and customization investments, the integrations that make a TMS the backbone are the same integrations that make leaving expensive.

That finding is the market's quiet economics. The TMS sits at the center of a web of connections, CMS, DAM, development environments, AI providers, translation engines, and every connection is an asset and a shackle. The scorecard rewards integration breadth, and the same breadth becomes the switching cost, which is why the report's governance guidance stresses that decentralization needs human oversight, the exit path needs to be designed while entering.

The workflow quote is the edition's thesis in one sentence: translations rely on a mix of machine translation, large language models, and human translation, a different blend for each touchpoint and language pair, and "the real uplift comes from workflow automation and agents, and TMS providers are not at all standardized in this respect."

The governance finding completes the picture. Pierce's guidance is that governance and decentralization must work in tandem: effective TMS platforms remove bottlenecks from decentralized execution, but human oversight is still required for standards, workflows, and performance monitoring. The phrase to carry is that complex organizations need skilled localization leadership, because no orchestration layer, however automated, replaces the judgment that decides what gets translated, by what engine, for which market, and what it must never say.

The honest limits of a first scorecard

A first edition has no predecessor, and its tier table defines the market as much as it ranks it. Three Leader placements are publicly confirmed, and the other nine placements have not surfaced. This article names what the record supports and nothing beyond it.

There is also the consolidation caveat. TransPerfect's Unbabel acquisition landed while the Wave was publishing, and the field's roster will not survive intact to the next edition. The scorecard is a snapshot taken during a market-shaping transaction, and its placements should be read with that timestamp.

Three questions for the localization buyer

Which pipeline is yours: fully automated, governed, or custom-model-plus-human? The three Leaders are three different products wearing one category name. Match the pipeline to your content risk, because regulated, high-volume, and creative content demand different engines.

What is your automation number, and what does it mean for your linguists? Ninety nine percent workflow automation is the edition's headline figure. Ask what the number is for your content mix and what the human review layer costs per word, because the automation number and the human layer are one budget.

What does the exit path look like, in writing? The report's own finding is that integrations lock customers in. Ask for the export contract, the glossary and translation memory portability, and the integration map before the first integration is built.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.

Worth reading alongside Localization Services: no Forrester Wave has ever scored this category, and the nearest evidence of what one would measure sits inside a translation technology evaluation that just got its own first scorecard instead.