Program Review · June 2026

The pilot that proved the model. Now we scale it across SciTechCONNECT.

A production NetworkOS deployment, configured for SciTechCONNECT, put DoW science & technology discovery in the hands of HBCU researchers — and they used it to co-author real proposals. This is what the pilot built, what it proved, and where it goes next.

Prepared forJason Preisser — Sr. VP of Federal Programs, ARI
Prepared byCollaboration.AI
EngagementSciTechCONNECT × NetworkOS pilot
0
cohort users onboarded
0
HBCU institutions live
0
data packs active
0
AI agents available
0
organizations in the network
Overview

What this review covers.

This is a comprehensive review of the SciTechCONNECT NetworkOS pilot undertaken by Collaboration.AI in partnership with the Applied Research Institute — intended as the basis for an executive review of next steps and expansion. For a program serving underrepresented research institutions, value is measured in access, discovery efficiency, and institutional capacity to compete for DoW S&T funding. The sections that follow document what was built, what it enables, and the evidence generated against those measures.

Objectives & status

The metrics we set — and where each one landed.

At kickoff, both teams acknowledged that downstream outcomes (awards, partnerships) are lagging indicators, hard to measure in a pilot window. Validation was therefore grounded in qualitative signal: engagement, feedback quality, and endorsement.

Objective
Status
Summary
Active platform usage
Achieved
29 users onboarded and engaged across structured sessions at all three institutions; sustained independent usage between sessions limited by HBCU summer availability — a known, confirmed dynamic.
Matches recognized as credible
Achieved
Cohort feedback confirms relevance and specificity of AI-generated results; users engaged at a sophisticated level, referencing specific tools and workflows.
Improved discovery efficiency
Directional
Strong qualitative signal that the platform surfaces opportunities users would not have found independently; a formal baseline comparison has not yet been run.
Endorsement letters
In progress
Sessions complete — three structured working sessions held through end of June; the team is now converting the outputs into endorsement letters.
Readiness for broader deployment
Achieved
Platform is stable, the delivery model is validated, and cohort engagement has produced sufficient signal to support an informed expansion decision.
Work completed

A fully operational deployment — built for SciTechCONNECT.

The pilot delivered NetworkOS as a production COTS platform, configured specifically for the program. Every core deliverable in the ROM is live.

Platform deployment & infrastructure

  • Production environment in commercial AWS, meeting security & availability requirements
  • Workspace built out and configured for ARI program office and HBCU cohort access
  • Admin layer for ARI staff — workspace management and usage visibility

AI agents, configured for the program

Two agents custom-built for SciTechCONNECT, plus two catalog agents available to the cohort.

  • Funding Opportunity Match — surfaces DoW S&T funding aligned to a research profile
  • Organization Capability Match — identifies partners aligned to capabilities
  • Catalog: Document Intelligence & Deep Research

A three-layer data environment

  • Public data packs — five live: Federal Hierarchy, SAM Entity, SAM Opportunity, SBIR, USASpending (Grants.gov in development)
  • HBCU Organization Library — a program-specific catalog, pre-seeded and enriched, now a living asset members build out themselves
  • Platform Organization Library — a cross-sector innovation network of ~19,000 organizations across industry, academia & government (detailed below)
Federal HierarchySAM EntitySAM OpportunitySBIRUSASpending

Cohort onboarding & engagement

  • 29 users across NC A&T, Winston-Salem State & Tougaloo — recruited with E4 Power
  • On-site sessions (NC A&T, WSSU) in April; virtual session for Tougaloo — full coverage
  • Between-session support: direct email, help center, and a self-service user guide
The wider network

A cross-sector innovation ecosystem — mapped and ready.

Beyond the program's own catalog, the platform's organization library maps a live innovation ecosystem — giving ARI a ready-made network for partnership matchmaking, tech scouting, consortium building, and regional economic development.

~19,000
organizations across industry, academia, and government at every level
Industry 55% Academia 33% Government 10%+

Anchored by institutions like MIT, Stanford, CMU, AFRL, national labs, and major primes — connected to thousands of emerging companies across ARI's priority domains:

BiotechAIMicroelectronicsAerospaceEnergyAdvanced manufacturing

Partnership matchmaking

Connect institutions with aligned industry and research partners across the ecosystem.

Tech scouting & consortium building

Find emerging companies and assemble teams around a capability or solicitation.

Regional economic development

See the innovation landscape by domain and geography to inform program strategy.

What it enables

Closing the access gap, in plain language.

The time, expertise, and infrastructure to pursue DoW S&T opportunities has historically favored larger institutions. The Funding Opportunity Match agent changes that dynamic — surfacing relevant solicitations matched to an institution's profile, not broad keyword search. Here is the agent working an illustrative HBCU research profile.

network.collaboration.ai/scitechconnect
NetworkOS — the My Network relationship graph
Terminal · Funding Match agent working

Illustrative example of the Funding Opportunity Match agent. Opportunity titles shown are representative, not live results.

Outcomes & value delivered

Where the cohort engaged, the platform delivered real work — not just search.

All core ROM deliverables are live: the configured workspace, all five data packs, and both custom agents in use. Beyond delivery, the pilot produced two durable assets — a living institutional library, and proof of the co-authoring workflow the platform was designed to enable.

Standout usage · co-authoring with an assistant
A single power user co-authored a federal grant proposal end to end over a multi-day working session — moving from funding call to collaborator to facilities and expertise alignment through successive proposal revisions, iterating with the agents on specifics like JSNN cleanroom capabilities and PDMS surface modification.
Funding call→ Collaborator→ Facilities & expertise→ Proposal revisions
Delivery against scope

Production-grade & complete

The workspace is provisioned, all five data packs are active, and both custom agents are deployed and in use. Stable and operating as designed.

A compounding asset

The HBCU library is alive

CAI pre-seeded the catalog from trusted sources; institutions now own their profiles and build them out — so matches sharpen as the program scales.

Engagement quality

Substantive, not exploratory

Members worked at a sophisticated level across funding discovery, collaborator matching, and proposal framing — treating NetworkOS as a serious working tool.

Lessons learned

What the pilot taught us.

01

Plan around the academic calendar

HBCU institutions are consistently less reachable in summer — a pattern E4 Power and ARI staff confirmed. Scheduling follow-on sessions after the April kickoff was difficult, which limited independent interim usage. Future cohorts should front-load structured sessions within the academic term.

02

Value concentrates in power users — by design

Engagement clustered around a single power user who drove the majority of activity. That is precisely the "co-authoring with an assistant" behavior the platform targets. Deployment should be built to find and enable these users, not measured by uniform usage.

03

Member-owned profiles are the moat

Pre-seeding got institutions to value on day one; letting them own and extend their profiles makes the library compound. The data asset is as important as the agents that read it.

04

Academic-first sequencing was sound

Focusing the initial phase on the HBCU cohort — deferring the industry pilot in coordination with ARI — gave the academic track the support it needed and produced credible signal. The choice has proven correct.

Feedback & product response

Engaged users shaping a platform they value.

Feedback was constructive and improvement-oriented — a cohort thinking in workstreams and research pipelines, not signaling friction. Most input is already in motion; the most strategic threads are directly shaping platform architecture.

What we heard
What we shipped / are building
Data freshness & sourcesQuestions on currency, an instance of outdated professional info, and requests for ORCID, Google Scholar, and institutional repositories.
Data packs carry timestamps and refresh on a defined schedule; the output-accuracy issue is resolved and in production. External source expansion is a top priority and underway. resolved + in progress
Funding source scopeSome members expected private-foundation and other non-federal funding to surface alongside DoW opportunities.
Scope is intentional — built around federal S&T funding aligned to the SciTechCONNECT mission. We'll set that expectation explicitly in onboarding framing, so users understand the boundary at entry. by design · onboarding framing
Profile & org managementThe manual burden of building profiles given faculty time, and a need to treat industry vs. academic as distinct record types.
Admins can now curate displayed attributes and distinguish org types; pre-population from external data and member self-contribution reduce the manual burden. shipped
Organizational hierarchyLabs, departments, and research centers operate as distinct entities — the flat org layer doesn't reflect how research actually works.
Institutional records with sub-organizations beneath them — sharper matching per research unit, plus a school-level rollup. In active development. in active development
Discovery → actionUsers finding solicitations wanted a direct path forward — inline links and context on the funding mechanism.
Closing the loop: inline opportunity links, curated resources, and funding-mechanism context. Shaping summer development sequencing. prioritized for summer
Clarify before a runRather than a wide, ambiguous prompt and a full agent run, users want the system to ask clarifying questions upfront to narrow intent.
Validates a capability already on the roadmap — multi-choice clarification prompts and conversational refinement before the agent runs. Cohort feedback sharpens the case to ship. on roadmap · validated
Workspace structure & privacyMembers organize by project and role and want artifact folders plus personal / institution / ecosystem visibility controls.
Among the most strategic input — directly shaping data segmentation and role-based visibility architecture. On deck. on deck
Query time & progress visibilityProcessing time on agent runs flagged as a meaningful concern in the final session.
In-app progress indicators and step-by-step status are already built — users watch work happen in real time and return to completed sessions without staying on screen. Feedback validated the need directly. built · validated
Support & onboardingClearer help paths, richer training, and a starter set of example prompts to help new users get going.
Help docs are live; expanded guidance is in development for full launch. A curated starter-prompt set is being built into the rollout onboarding package — a program-level deliverable. live + expanding
Endorsement & evidence

Driving to the evidence the continuation decision needs.

The primary evidence artifact is endorsement letters from participating institutions — direct attestation from the people who used the platform. The three-session working series is now complete; the team is converting the outputs and feedback collected across all three sessions into completed endorsement letters.

April 2026

On-site & virtual onboarding complete

NC A&T and Winston-Salem State on-site; Tougaloo virtual. 29 users onboarded across the cohort.

June 18, 2026

Working session 1 — evidence series begins

First of three structured sessions designed to answer key evaluation questions directly.

June 30, 2026

Final working session — series complete

Concluding session with NC A&T and Tougaloo attending together; the structured series is now closed out.

Now

Converting outputs into endorsement letters

The team is turning session outputs and feedback into completed endorsement letters for the continuation decision.

Recommendations

Future enhancements & capability expansion.

The pilot's feedback gives a clear, validated roadmap. Each of these is already scoped — several in active development.

Organizational hierarchy

Sub-organizations (labs, departments, centers) beneath an institutional record — sharper matching and a school-level rollup. In active development.

External source expansion

Connect richer research signal — ORCID, Google Scholar, institutional repositories, grant platforms — to deepen profiles and matches.

Discovery-to-action loop

Inline opportunity links, funding-mechanism context, and curated next-step resources — plus clarifying questions before a run — so discovery converts to pursuit inside the platform.

Workspace & role-based visibility

Project-based grouping and artifact folders, with personal / institution / ecosystem access controls that map to how the program operates.

Reduced profile burden

Pre-populate profiles from trusted external data and let members contribute directly — lowering faculty effort while improving accuracy.

Baseline efficiency study

Run a formal discovery-efficiency comparison against manual methods to convert today's strong directional signal into a hard metric.

Deployment & scaling

The recommended path to full rollout.

Academic-first sequencing has proven sound — and the pilot surfaced a clearer picture of what full rollout should look like. Our recommendation: commit the next phase fully to the HBCU academic rollout, with the marketing, training, and onboarding it needs, before bringing industry on as active users.

Now

Commit to the academic rollout

Marketing, training, and onboarding built for a time-constrained cohort — so adoption is smooth and the value proposition is obvious.

Jul 30–31

HBCU funding symposium

The DoW's HBCU-focused symposium (hosted by Evelyn Kent) — the anchor for an amplified outreach push. An amplification milestone, not a go-live date.

Apr 2027

Confirmed runway to full rollout

Room to sequence deliberately without risking the timeline; the full academic experience follows the symposium at a pace that protects quality.

Sequencing, not scope reduction. Industry partners remain represented in the platform today as organizations and resources in the matching environment — so cross-sector discovery isn't lost during this phase. They are simply not yet onboarded as direct users.

Open item — the ~1,200-user figure. The ROM scoped ~1,200 users across academia and industry from membership counts at the time. With academia now the near-term priority and HBCU enrollment growing since, that number warrants a fresh look — align on scope, user count, and how growth is managed for cost and access before full rollout begins.

Foundation for expansion: independent of how these resolve, the pilot's configuration — data packs, AI agents, workspace setup, and the HBCU library — is a repeatable template that significantly reduces setup time for every new cohort.

The bottom line

The pilot proved the model. The cohort proved the value.

A production platform, a living institutional asset, and researchers co-authoring real proposals — built on a template that's ready to scale across SciTechCONNECT. Let's align on sequencing at the review.