Collaboration and Innovation

The pitch is everywhere now. Partner or die. No single institution can build modern technology alone. Open innovation. Ecosystems. Co-opetition. Strategic alliances. Every consulting deck and trade publication reaches the same conclusion: the future belongs to organizations that collaborate well.

I have spent twenty plus years inside Canadian credit unions, in VP roles spanning risk, IT, innovation, and operational excellence. I have served as a director on boards, participated as a Filene i3 cohort member, contributed as a subject matter expert and team lead on core banking and digital conversions, delivered some of the "firsts" on digital banking, and helped deliver Canada's first contactless payment wristband and the first online membership application in the credit union space. Most of those wins were collaborations. Some of the failures I have lived through were also collaborations.

Here is what I have learned: the partner or die framing is not wrong, but it is nuanced. It reframes a difficult discipline as if it were a default. A CMO Council survey of 330 senior executives found that while 85 percent of companies view partnerships as essential, almost half report failure rates of 60 percent or higher. Other data states that over 70 percent of partnerships fail before they ever enter a paid pilot, and average conversion from proof of concept to scaled deployment is roughly 25 percent. For AI specifically, IDC (International Data Corporation) and Lenovo report that 88 percent of proofs of concept never reach widescale deployment and that only 16 percent of AI initiatives achieve scale beyond the pilot stage. The problems in every one of these studies cluster around governance, risk management, infrastructure readiness, change management, and executive ownership.

Filene, the most credible neutral voice in the credit union sector, puts it more diplomatically. From their own collaboration research: "Credit union leaders haven't been convinced of the value of collaboration." That is a remarkable thing for an institution founded on cooperative principles to write.

Pattern of Failure 1: Structure determines whether value reaches members

The structural promise of cooperative collaboration is also its structural risk. When five or fifteen credit unions sit at a table to decide what needs to get built, the answer that survives is the one no one objects to. Typically the result is a lackluster solution that does not create any differentiation in the marketplace or for members.

I have watched this play out in collaboration governance meetings and in innovation cohorts. The conversation starts with ambition. Within a couple of meetings, every meaningful design choice has been softened to accommodate the most risk-averse participant. The output usually ships. It usually works, and it is indistinguishable from what a third-party vendor would have sold each institution separately. The shared cost is real. The shared advantage is not. CEOs, rightfully so, question if it's worth their people's time and effort.

Each credit union wants the group to produce something proprietary enough to create value, and inclusive enough that no one is left out. Those two requirements do not play well together. The cooperatives that scale innovation do not pursue consensus. They pursue clear ownership of differentiating layers, shared investment only on commodity layers, and explicit governance that holds the service organization accountable to outcomes.

McKinsey's recent work on sovereign AI ecosystems calls this principle minimum sufficient sovereignty. Classify workloads by the importance of regulatory issues and third-party exposure, then assign each layer explicit requirements for data residency, key ownership, and access controls.

Translation: Collaborate on the plumbing. Own the member experience. More on this in a future article.

Pattern of Failure 2: Without a translator, the partnership is dead before it starts

The Harvard Business Review's March 2026 issue captures something I have participated in for years but never had a clean name for, which has led to many successful innovation and collaborative efforts. Linda Hill and her co-authors argue that innovation increasingly depends on partnerships, and the leaders who make them work are bridgers, who excel at curating partners, translating across their ways of working, and integrating efforts to maintain momentum. They are the glue that makes it work.

Bridgers are not always the most senior people in the room. They are rarely the loudest. They are the ones who can sit in a fintech founder's pitch meeting in the morning, a credit union risk committee in the afternoon, and a board meeting in the evening, and faithfully represent each group to the other. They speak multiple dialects of the same language and can move between them, representing without compromise. They understand that the fintech's product roadmap and the credit union's regulatory examination cycle are not just different timelines but completely different worlds.

In my experience, the partnerships that scaled had at least one clear bridger. The ones that died usually died because the bridger left, was reassigned, or was never identified in the first place. A vendor relationship managed by procurement is not a partnership. A fintech relationship managed by a credit union innovation team that has never sat with a member services rep is not a partnership either.

This is also why some of the strongest credit union collaborations I have seen were not formal cooperative structures at all. They were small, focused partnerships between two or three institutions with executives who had worked together for years and trusted each other's instincts. The legal structure followed the relationship. Not the other way around.

Pattern of Failure 3: Pilot purgatory is a process problem, not an idea problem

The most consistent finding across the AI scaling research is also the one most relevant to credit union innovation today. Across four independent industry analyses (MIT, IDC, IBM, EPAM), 80 to 95 percent of AI proofs of concept fail to reach production or fail to deliver measurable P&L impact. The diagnosed causes converge on a small set of organizational variables: governance, infrastructure readiness, change management, executive ownership, and the depth of workflow integration. The MIT analysis found that vendor-led, workflow-integrated implementations succeed at roughly twice the rate of internal builds. The variable that predicts success is not the model, the team, or the budget. It is the depth of workflow integration.

This is all too real. A vendor demo that solved a real problem in a controlled environment. An enthusiastic pilot in one branch or one product line. A six-month delay before integration with the core. A scope expansion that exposed legacy data quality the pilot never touched. A change management plan that arrived three weeks before go-live. Eventually, a quiet abandonment of the project, or a change management nightmare.

Every documented failure mode in collaborative innovation, whether the artifact is a fintech partnership, a shared services platform, or an AI pilot, is a continuous improvement problem. No defined value stream. No clear flow from pilot to production. No pull from the front line. No measurable outcome defined at the start. No respect for people in the change rollout. No standard for what "ready to scale" means before launch.

The credit unions and mid-sized institutions that scale innovation through collaboration are not the ones with the best partnerships. They are the ones with the discipline to absorb what those partnerships produce. The lean and continuous improvement practitioners I know would call this absorptive capacity (Muda, Mura and Muri) in their own vocabulary. The popular discourse on collaboration treats partnership access as the constraint. Find the right partner, do the right deal, and value flows. The empirical evidence says the constraint is somewhere else: in the receiving organization's capacity to absorb what the partnership produces.

Are you ready to collaborate?

So when should a credit union or any mid-sized regulated institution actually pursue collaboration as the path to scale?

Not as a default. Not because everyone else is doing it. Not because the trade press says so. Specifically: when the capability in question is genuinely commodity at the layer where you are collaborating, when you have a credible bridger on your side of the table, when the partnership is wired into a workflow rather than handed off arms-length, when the governance of any shared service organization holds it accountable to member outcomes rather than to its own institutional growth, and when your internal operating discipline can absorb whatever the partnership produces.

Outside those conditions, the data and my experience converge on the same answer. Collaboration is more likely to drain energy than to scale anything.

What this means for the challenges ahead

The partner-or-die rhetoric is not wrong about the underlying shift. AI, cloud, digital banking, real-time-rails, open-banking and and regulatory complexity have raised the cost of going alone past what most mid-sized institutions can afford. But the conclusion most consultants draw, that the answer is more partnerships and bigger ecosystems, misses the actual constraint.

The constraint is not access to partners. It is the discipline to make partnerships produce something worth scaling. That discipline has a name. The credit union sector and the broader regulated-industry world have spent decades developing it under labels like lean, continuous improvement, and operational excellence. The next decade of credit union innovation will reward the institutions that recognize they already have the operating system. They just have not been pointing it at innovation collaboration.

That is the work. It is unglamorous. It does not show up on a vendor's slide deck. It is what twenty years inside the credit union system has taught me.

Now, you know the problem. And as I am sure I read somewhere, when you are able to identify and articulate the problem, you are 80% there.