Act A — The Market That Isn’t Working
There is a moment that employment support workers know well. They call it “the tour.”
You bring a client — someone with a developmental disability or autism spectrum condition — to visit a potential employer. The employer is, in principle, enthusiastic. They have signed the inclusivity pledge. They have attended the diversity training. They want this to work.
Then the tour starts, and the employer begins filling the silence with the wrong jobs.
Warehouse sorting. Breakroom cleaning. Paper shredding.
The client says nothing. The employment worker says nothing. They all walk through the facility being polite, and nothing comes of it — not because the employer is bad or the client has nothing to offer, but because nobody in that room has the vocabulary to describe what the client actually can do. Nobody has given either party a neutral, specific starting point for a different kind of conversation.
Inclusive employment, in 2026, still largely works this way. In theory, the market exists: employers with genuine operational needs, job candidates with specific and real capabilities, and a professional support system designed to connect them. In practice, the market is thinner than it appears. The tools available — job boards, agency intake forms, warm introductions — are built for a different kind of matching. They are built for people who can describe themselves fluently in the language of the job market, and for employers who know how to specify what they actually need.
When both sides of a potential match are, in different ways, unable to describe the deal they are looking for, the market fails quietly. Nobody lists this as a missed transaction. It simply never happens.
The numbers are not abstract. The Employment Accessibility Research Network estimates that fewer than one in three working-age Canadians with significant disabilities is employed, despite the fact that a substantial proportion could work — and work productively, reliably, and at skill — in contexts designed to accommodate their capabilities. Inclusive employment agency staff are skilled and committed professionals. They are not the bottleneck. The bottleneck is the matching infrastructure.
What follows is a fictional illustration of what a different kind of matching infrastructure might look like.
Act B — The Story
The Problem with Demarco
Demarco Nkosi was twenty-four years old and very good at something that had no name in any job description he had ever read.
He had grown up in Brampton, Ontario, the son of a Ghanaian father who worked cargo logistics and a mother who managed a medical supply warehouse in Mississauga. He’d spent hours as a teenager watching the warehouse’s automated pill counter make errors that no one caught until the daily reconciliation. He started keeping his own log. Within six months, he was finding counting errors that the machine registered as complete — not by algorithm but by memory: he knew what a correct blister pack looked like from the orientation of the foil, the slight offset in the embossing that indicated a wrongly-seated tablet.
He had a name for it. He called it “knowing wrong from not-quite-right.”
His autism spectrum diagnosis had come at thirteen, his ADHD diagnosis at fifteen. The school system had given him an educational assistant and a quiet room for exams. The post-secondary system had given him a disability services coordinator and an extra semester. The employment system had given him three years of supported employment at the sheltered workshop program in his neighborhood, where he assembled promotional materials and packaged fundraising kits — tasks so far below his capability that he stopped talking at work, because there was nothing worth saying.
His employment support worker, Patience Asante, was the one who noticed. She’d watched him catch a misprint on a batch of packaging that two other staff had already reviewed and signed off. She noted it in his file: detailed visual inspection, pattern anomaly detection, sustained concentration on repetitive tasks.
She had no idea what to do with that observation. The tool she used to match clients with employers was a general job-board integration. It searched open postings for accommodation-friendly workplaces. What it could not search for was: employers who need someone who can reliably detect visual pattern anomalies in printed pharmaceutical packaging and sustain that concentration across an eight-hour shift.
That job posting does not exist. The employer who needs it doesn’t know they’re looking for it.
The KnowledgeSlot Layer
The Inclusive Employment Exchange — a pilot platform deployed by a consortium of provincial employment support agencies and two Toronto-area university programs — was not built on a job board.
It was built on a matching engine that started from the supply side rather than the demand side.
The platform’s KnowledgeSlot layer held a curated knowledge base assembled by occupational therapists, job coaches, and disability employment specialists over eighteen months. It contained not job descriptions but work function taxonomies — granular catalogues of the specific subskills embedded in industrial, commercial, and services roles:
- sustained repetitive visual inspection
- fine motor repetition with quality gate checkpoints
- multi-step documented procedure compliance without verbal supervision
- inventory counting against a known reference
- sensory environment tolerance (noise, lighting, smell, temperature)
Each function was mapped to accommodation requirements, environmental specifications, and — critically — to industries and roles where that function generated measurable commercial value.
Visual pattern anomaly detection in repetitive printed materials: pharmaceutical blister pack inspection, label verification, surgical device packaging compliance, banknote quality control.
When Patience onboarded Demarco into the platform, she did not fill out a job application. She had a conversation — partly in the platform’s interface, partly over WhatsApp where Demarco was more comfortable. The AI asked about what he found easy, what he noticed that others didn’t, what kinds of environments he had worked in and which had felt sustainable. It asked Patience about the performance she had observed: the misprint catch, the reconciliation log, the warehouse hours.
Demarco’s profile, when the platform assembled it from that conversation, was not a CV. It was a capability map — detailed, specific, structured around work functions rather than job titles.
The Demand Side Nobody Had Mapped
Two hours east of Brampton, in a clean-room packaging facility on the outskirts of Guelph, Ontario, a contract pharmaceutical packager called Meridian PharmaPack was quietly managing a problem.
Meridian’s quality manager, Adriana Kowalczyk, was three inspectors short on her blister pack visual inspection line. The role was essential — regulatory requirement, not preference. Health Canada’s Good Manufacturing Practice guidelines required documented visual inspection at the secondary packaging stage. But finding people who could do it well was surprisingly difficult.
The work required sustained, quiet concentration across long shifts. It required the ability to notice a 0.3mm foil misalignment or a tablet orientation error against a known-correct reference. Most people found it intolerable within two weeks. The ones who found it natural — who could sustain it for eight hours and finish a shift not exhausted but satisfied — were genuinely rare.
Adriana had run the role through two staffing agencies and received people who lasted days. She’d tried incentive bonuses and improved ergonomics. The role still sat vacant.
When Meridian was approached by a regional employment support network about participating in the Inclusive Employment Exchange pilot, Adriana agreed with limited expectations. She had done the accessibility workshop. She had done the tour. She was skeptical.
The platform’s intake was different. Instead of asking her to post a job description, it asked her to describe what the role actually required — in operational terms, not HR terms. An occupational assessment module walked through the work environment: noise levels, lighting, standing versus sitting, supervision model, shift structure, quality gate frequency. It asked about the consequences of error and the cadence of feedback.
Meridian’s posting on the platform was not a job description. It was a work environment profile — specific, structured, and matched to the KnowledgeSlot’s function taxonomy.
The Match
The semantic matching engine in Cosolvent — the platform’s deal-assembly harness — did not match Demarco to Meridian PharmaPack because of a keyword overlap between their profiles.
It matched them because Demarco’s capability map aligned with Meridian’s work function requirements across six specific dimensions: visual inspection pattern recognition, sustained repetitive task tolerance, fine motor accuracy, documented procedure compliance, low-verbal-supervision preference, and preference for structured environmental conditions.
Neither Demarco’s profile nor Meridian’s posting had used the word “inspection” in the same field. The match was semantic — the platform understood that what Demarco did when he kept his reconciliation log at his mother’s warehouse was structurally the same work function Adriana needed filled on her blister pack line.
Both parties received match notifications the same afternoon.
Demarco’s notification came through WhatsApp. It described Meridian in plain language — what they made, where they were, what the shift looked like — and noted that the platform had found a specific alignment between his documented capabilities and a real operational need. Not a test. Not a tryout. A match based on what he actually did.
Adriana’s notification came through email. Meridian had been matched with a candidate whose capability profile — assembled by a certified employment support specialist — documented a specific operational fit with the secondary packaging inspection function.
With the match came a document neither party had requested.
The Generative Match Story
The platform generated what it called a Match Story — a plain-language narrative of how this specific engagement could work, assembled from three sources: Demarco’s capability profile, Meridian’s work environment profile, and the KnowledgeSlot’s curated knowledge of supported employment best practices in the pharmaceutical manufacturing sector.
The Match Story described a twelve-week supported employment trial structured around Health Canada’s GMP inspection documentation requirements. It named the accommodation adjustments that were standard in this setting: a dedicated inspection station with consistent lighting, a visual reference card for known-correct configurations, a documented daily quality gate checklist, and a job coach check-in cadence that reduced verbal supervisory contact to structured end-of-shift reviews.
It identified two regulatory elements relevant to Meridian: the GMP requirement for documented visual inspection competency assessment, and the Ontario Disability Support Program’s Supported Employment wage supplement — a funding mechanism that offset a portion of employer onboarding cost during the trial period. Adriana had not known the ODSP supplement existed. Meridian’s HR team had not known GMP documentation requirements could be satisfied by a structured supported employment protocol.
The Match Story was not a contract. It described one way the engagement could work — a hypothesis, grounded in the specific profiles of both parties and the regulatory context of the sector.
Adriana read it twice. The second time, she forwarded it to Meridian’s compliance officer with a single line: “Is this actually compliant with our GMP audit requirements?”
The compliance officer confirmed it was. More than that — the structured daily quality gate checklist described in the Match Story was better documentation than what the previous inspectors had been asked to produce.
Demarco read his version of the Match Story with Patience. The job coach model it described — structured, low-verbal, with explicit feedback at the end of each shift — was exactly the supervision environment in which he’d performed best. He asked Patience one question: “Does the ODSP supplement mean they’re being paid to hire me, or that I’m worth less to them?”
Patience explained how it worked. His first message to the platform, sent through WhatsApp, was direct: “Can you confirm they’re paying the full rate from week one, with the supplement covering something separate?”
Meridian confirmed within the hour.
The Deal
The twelve-week trial began on a Monday in March. Demarco took the GO train from Brampton to Guelph, a commute he had memorized in detail the week before.
By week three, his inspection error-catch rate was higher than any of the three inspectors Adriana had previously employed. He was catching foil misalignments at a rate that exceeded the line’s formal quality gate target. He did not find this remarkable; he found it interesting.
By week eight, Adriana had extended the offer to a full-time permanent position. The job coach’s formal involvement ended at week twelve, as planned. A platform-generated transition brief documented what accommodations remained in place and what Meridian’s internal supervisor needed to maintain them — a document structured around the GMP audit trail Meridian was already keeping.
Demarco would later tell an employment support conference that the thing he remembered most about the process was the Match Story. “It said what the job actually was,” he said. “Not what they needed someone to be, but what the work actually required. It matched that to what I can actually do. I’d never read something like that before.”
Act C — The Structural Reading
This story is fictional. Demarco Nkosi and Meridian PharmaPack do not exist. But the market failure they illustrate is real.
The disability employment market in Canada and in most OECD-comparable countries shows the structural signature of a thin market: structural desire to exchange on both sides — employers with genuine operational needs, job-ready candidates with specific and real capabilities — and a persistent failure to match. Not because the participants are unwilling, not because the jobs don’t exist, but because the matching infrastructure is built for a different kind of transaction.
The forces are identifiable:
Strategic information withholding, dual-sided. Employers don’t post the specific operational parameters of their hard-to-fill roles because they don’t believe an employment support agency will understand them. Workers with disabilities and their advocates don’t disclose specific capability profiles because the typical onboarding process punishes over-disclosure — you end up sorted into the lowest-skill roles regardless of what you reveal. Both sides play down, and both sides lose.
Opacity through vocabulary mismatch. The language of disability employment support — accommodation needs, supported employment, job coaches — is not the language of production floor operations — inspection frequency, error rate thresholds, GMP compliance, shift structure. Neither party can search for the other because they don’t share a vocabulary. The KnowledgeSlot’s work function taxonomy is the bridge that existing tools don’t build.
Cognitive overload on both sides. An employer managing three vacancy crises simultaneously cannot attend specialist employment fairs and learn the regulatory landscape of supported employment at the same time. A job candidate with executive function challenges cannot navigate the job market’s standard onboarding requirements while simultaneously managing accommodation documentation requirements. The platform reduces cognitive load by doing the translation work — and by starting the conversation with a concrete, specific Match Story rather than an open-ended application.
Regulatory fog. The ODSP employer wage supplement exists. The GMP documentation flexibility exists. Neither is well-known to the other side of the market. The KnowledgeSlot turns regulatory fog into actionable information for both parties.
Participant scarcity, asymmetric. The supply of employers genuinely equipped for inclusive employment is truly small — not because employers are unwilling in principle, but because the infrastructure to support the employment relationship is missing and the cost of building it ad hoc is prohibitive. The platform provides that infrastructure at a fraction of the per-engagement cost of a bespoke agency integration.
This is not primarily an accessibility story. It is a market design story. Demarco is not a charitable case — he’s a specialist. The same AI-driven innovations that are transforming commodity trade and cross-border professional services can be applied to employment markets where the matching vocabulary is missing and the structural information asymmetry is deep.
MarketForge’s sponsor model suggests a natural institutional home for this kind of platform: a provincial ministry of labour, an employment support agency network, or a university-affiliated social innovation incubator — organizations that can curate the KnowledgeSlot’s domain knowledge, carry the trusted intermediary role, and operate the platform as a public or social benefit rather than a pure commercial service.
The technology exists. The structural desire exists. The gap is the matching infrastructure.
The characters and organizations in this story are fictional. Any resemblance to real persons or entities is coincidental. The regulatory frameworks referenced — Health Canada GMP guidelines, ODSP employment supports, Ontario’s Accessibility for Ontarians with Disabilities Act — are real and are accurately characterized above.
DeeperPoint is developing the open-source tools that underpin platforms like the one described here. Learn more about the MarketForge platform, the thin market problem, and who these tools are designed for.