Multiverse

Customer Story

How one engineer ran Multiverse's entire accessibility program, end to end.

Multiverse needed a VPAT to compete for government contracts. With a product spanning 13 user-facing domains built by six engineering squads, the manual path was projected at a year of work. With Workback, a single engineer ran audit, remediation, and certification, and the finished VPAT now rates 45 of 50 WCAG criteria as Supports.

402 merged pull requests across 16 repositories. One engineer.

About Multiverse

Multiverse is the UK's leading EdTech platform for apprenticeships and applied training. Its programs in AI, data, engineering, and leadership serve more than 30,000 learners across 1,500 customers, delivered through human coaching plus Atlas, a Socratic AI coach, and measured on business outcomes. The company exists to open access to opportunity, and that mission only holds if the platform itself is usable by everyone who depends on it.

Multiverse's scale makes that hard in a specific way. The product spans 13 user-facing domains, built by six independent engineering squads and roughly sixty developers. Accessibility at that scale is not one fix in one codebase. It is hundreds of fixes spread across teams that each have their own roadmap.

Headquarters
London, UK
Learners
30,000+
Engineering squads
6
User-facing domains
13

The challenge

A VPAT to win contracts, and no year to spare.

In the UK, the Equality Act 2010 places a duty on organizations to give disabled people equal access to digital services. Because apprenticeships are delivered with public funding, Multiverse sits closer to the stricter public sector bar, where WCAG 2.2 AA is the benchmark, than the private sector floor. Multiverse chose to meet that standard on its own terms, ahead of being made to, because it is the right thing for its learners.

There was also a commercial clock. Multiverse was pursuing a new slate of government contracts that required proof of conformance. The team knew accessibility was baked into how they build. What they did not have was a VPAT to prove it.

The manual path did not fit the clock. A third-party audit alone was projected at two business quarters, remediation across six squads at a year, and twelve developers were slated for the effort. An upcoming brand refresh gave the team a natural moment to get ahead of it, with a clean before and after to measure against. Kate Pincott, Multiverse's Director of Design, tied the program to the refresh, set the mandate that impact be measurable, and directed recurring fixes through the design system so a single correction would compound across every squad.

The audit

Weeks, not quarters.

With Workback in place, Alexander Langolf, a senior frontend engineer at Multiverse, ran the audit single-handedly. The twelve developers originally slated for the program stayed on their roadmaps.

Workback's audit agent, Ada, drove sessions end to end across Multiverse's authenticated product the same way real users move through it: as learners, coaches, employers, and administrators. It handled live authentication, cookies, and sessions, and built a working map of the entire product, so that for the first time the team knew exactly where issues lived and which squad owned them.

Every finding then went through human review before reaching an engineer. Of 688 issues flagged, 577 were confirmed as real; the false positives and duplicates were screened out before they could cost anyone time.

56
User flows audited

Across five distinct actor types, from learners to administrators.

577
Confirmed issues

232 critical, each mapped to the repository and squad that owns it.

Days
Not quarters

The audit that was projected at two business quarters took days of active work.

"I can't believe I was able to drive the entire accessibility program, from audit to remediation across all of engineering, basically by myself with Workback."

Alexander Langolf, Senior Frontend Engineer, Multiverse

Remediation

Fixes that ship in the workflow developers already use.

With a manual audit, this is where momentum stalls: a spreadsheet of findings, weeks of meetings to assign owners and argue priorities. Workback moves straight from finding an issue to fixing it. Its GitHub App opens pull requests directly in Multiverse's repositories, so fixes arrive where engineers already work and get reviewed in code review like any other change.

The question every engineering leader asks first is whether those PRs are good enough to merge without rework, or whether they just create a review burden. At Multiverse, 93 percent of published PRs merged, with a median time from open to merge of three days.

Where the same issue repeated across squads, such as focus states, contrast, and labelling patterns, the fix was routed through the design system instead of patched repository by repository. One corrected component propagates to every product that consumes it. That is how 577 issues becomes a tractable number: a meaningful share of them share a root cause, and the root cause gets fixed once.

402
Pull requests merged

93% of the 432 PRs Workback opened. Remediation at scale, not one fix at a time.

16
Repositories

Fixes routed to the squad that owns each repo, across all six squads.

3 days
Median time to merge

PRs reviewed and merged inside each squad's normal cadence.

"I like thinking about Workback as the best accessibility linter with a --fix flag for the Multiverse platform."

Alexander Langolf, Senior Frontend Engineer, Multiverse

The VPAT

A living record, not a static report.

The VPAT was the reason Multiverse started. By the time remediation wrapped, it was already done: Workback generates the VPAT from current findings and updates it as fixes land, so the report Multiverse hands to a procurement team reflects the product as it is today, not as it was the quarter an auditor last looked.

The progression is the before and after leadership asked for. At the June baseline, 20 of 50 WCAG 2.1 AA criteria rated Supports across the learner experience. By September, the finished report rated 45 Supports and 5 Partially Supports, with zero criteria at Does Not Support.

June 2026
20 / 50

Criteria rated Supports at the baseline audit.

August 2026
27 / 50

Mid-program, with PRs merging across squads.

September 2026
45 / 50

Final report: 45 Supports, 5 Partially Supports, zero Does Not Support.

"Going from audit to merged fixes in weeks meant accessibility stopped being a backlog item and became part of how the product ships."

Kate Pincott, Director of Design, Multiverse

Results

What changed.

45/50

WCAG 2.1 AA criteria rated Supports on the final learner platform VPAT, up from 20 at the baseline. Zero Does Not Support.

402

Pull requests merged across 16 repositories and six squads, a 93% merge rate.

515

Issues closed, with 403 fixes verified live in production.

1

Engineer running the program, instead of the twelve developers originally slated.

~2 wks

Of hands-on engineering effort across the whole program, audit to certification.

description

A VPAT that updates itself as fixes land, instead of a static report that ages the day it is delivered.

"The compounding effect is what mattered most: route a recurring fix through the design system once, and every squad inherits it for free. That's the difference between chasing compliance and designing it in."

Kate Pincott, Director of Design, Multiverse

The deadline doesn't move. Your roadmap shouldn't either.

Your engineers want to build product, not chase compliance. Talk to us about running a pilot before your next district renewal.

Start Your Free Pilot