Innovación Abierta

Por Clara

Última actualización

Open innovation is a model in which firms deliberately manage knowledge flows across organizational boundaries (bringing external ideas in and moving internal knowledge out) to accelerate their R&D and commercialize technology more effectively than they could alone. Chesbrough coined the term in 2003, and the later Bogers-Chesbrough formulation became the field's working definition: "a distributed innovation process based on purposively managed knowledge flows across organizational boundaries." The word purposively is doing real work there. Open innovation is not an accident of collaboration. It requires deliberate architecture.

Most corporate programs built under this label do something simpler: they scan startups, run an innovation challenge, and call it open innovation. That narrower practice is inbound-only, and it is why so many programs generate activity without generating value. Readers comparing it with invention, the innovator's dilemma, or the wider innovation glossary should keep the governance layer in view.

TL;DR

• Open innovation means deliberately managing knowledge flows in two directions: inbound (outside-in) and outbound (inside-out), often combined in coupled partnerships.
• Inbound, outbound, and coupled programs do not share the same IP requirements. Most programs fail when they are designed before anyone specifies which variant is in use. See §4 for the variant-by-variant IP governance breakdown.
• IP governance is not a legal afterthought. A 2025 study of Belgian firms found that broader collaboration portfolios were associated with higher abandonment and lower completion rates, which turns governance into a design question, not just a legal one.
• The empirical relationship between openness and performance is an inverted U, not a straight line. Both too little and too much openness underperform, as Laursen & Salter (2006) showed.
• Open innovation is the wrong choice when you cannot protect the value you disclose, when your firm lacks the internal capacity to absorb external knowledge, or when the collaboration overhead exceeds the knowledge you expect to gain.

Open innovation requires a governance architecture that most corporate programs skip. The inbound-only trap — running partner scans and idea challenges without building outbound or coupled flows — is the single most common reason programs generate activity without generating value. Firms that get the most from open innovation treat IP governance as a design input, not a legal afterthought.

What is open innovation?

Open innovation is a model in which firms treat knowledge as something that flows across their organizational boundaries in both directions, rather than something produced and kept entirely in-house. The defining feature is purposive management of those flows, which distinguishes it from generic partnerships or vendor relationships.

Chesbrough coined the term in 2003 to distinguish firms that use external ideas and external paths to market alongside internal R&D. The 2018 Bogers-Chesbrough-Moedas refinement sharpened that framing into the field's working definition: "a distributed innovation process based on purposively managed knowledge flows across organizational boundaries." That definition makes the management layer explicit. The flows matter, but so do the rules that decide what can move, who can use it, and how value is captured once it does.

Gassmann and Enkel (2004) added the taxonomic precision that practitioners and researchers have relied on since, formalizing three archetypes: inbound (outside-in), outbound (inside-out), and coupled. Gassmann & Enkel (2004) Chesbrough's 2003 work introduced the open/closed paradigm. Gassmann and Enkel named the structural variants that make it actionable.

What open innovation is not: R&D outsourcing, vendor management, or the wisdom of the crowd. Each of those is a specific mechanism that may or may not constitute open innovation depending on whether knowledge flows are being purposively managed across organizational boundaries.

How did open innovation emerge — and what did 20 years prove?

What changed after 2003 was the managerial claim that firms should design a repeatable way to find, license, integrate, and commercialize outside ideas. The adoption curve reached scale in large firms, but Chesbrough's 20-year retrospective found that most programs skewed toward inbound-only mode, with outbound and coupled flows underused and failure conditions understated.

That imbalance shows up in the evidence. Chesbrough's later commentary is unusually candid about how often open innovation research overweights success stories and understates failure conditions. The newer abandonment data points the same way: collaboration breadth rises faster than program infrastructure, so more partner types can mean more hold-ups and lower completion rates instead of faster learning.

The adoption curve did reach scale in large firms. Chesbrough and Brunswicker's 2015 survey found that OI practices were widespread across large multinational corporations. Inbound practices still dominated, while outbound and coupled modes remained underused relative to their theoretical importance. Collaboration also became routine at the macro level. In 2026, the OECD reported that 55% of enterprises across member countries were innovation-active, accounting for 68% of employment.

What are the three structural variants of open innovation?

The three variants are not a spectrum of openness. They are structurally distinct models with different IP requirements. Treating all three as interchangeable under the "open innovation" label is how programs get designed for the wrong variant.

Source: Gassmann & Enkel (2004). West & Bogers (2014)

Inbound open innovation (outside-in)

Inbound OI means enriching the firm's knowledge base by integrating external sources such as suppliers and universities, along with other outside contributors. This is the dominant mode in practice. LEGO Ideas is a well-known consumer-facing variant: fans submit product concepts, community members vote, and proposals exceeding 10,000 supporters trigger a formal LEGO review. By 2019, the platform had received more than 26,000 submissions, of which 23 became commercial products. The 1% royalty paid to winning creators is a defined IP transfer mechanism, evidence that even consumer-facing inbound programs require explicit IP structure.

Outbound open innovation (inside-out)

Outbound OI involves moving internal IP and knowledge outward, licensing, spinning out, releasing patents, or publishing know-how to generate revenue or grow an ecosystem. Tesla's 2014 patent pledge used outbound to build EV ecosystem adoption. Qualcomm's licensing division uses outbound to generate revenue from mobile standards IP at a 72.4% margin.

Coupled open innovation

Coupled OI links inbound and outbound through alliances, joint ventures, or pre-competitive consortia, where firms work with complementary partners and knowledge flows in both directions under a shared governance agreement. Semiconductor firm IMEC and the pharma Pistoia Alliance are canonical examples (detailed in §8). Coupled OI requires the most governance infrastructure of the three variants and the most explicit pre-competitive boundary decisions about what is shared versus withheld.

Why does IP governance determine whether open innovation works?

Corporate open innovation mostly fails because companies add partners faster than they add transfer rules, making knowledge harder to license, route, and commercialize. A 2025 peer-reviewed study of Belgian firms found that broader collaboration portfolios produced worse project completion rates, not better ones — confirming that governance architecture, not partner volume, determines program value.

The evidence for this claim is quantitative. Van Criekingen, Freel, and Czarnitzki found that open innovation leads to higher rates of project abandonment and lower rates of project completion. Belgian firms working with six or more collaborator types completed just 44% of their innovation projects, versus 60% for firms that collaborated with no one. Firms collaborating with four to five partner types landed at roughly 50%.

That study does not measure IP-governance quality directly. It does show the cost of adding relationship complexity without adding enough structure to keep ownership, licensing, and commercialization decisions legible. The completion-rate collapse from 60% to 44% is not an argument against collaboration itself. It is an argument against unmanaged collaboration.

Belgian firms using six or more collaborator types completed just 44% of innovation projects, versus 60% for firms with no external collaboration.

IP governance by variant

The specific governance requirements differ by variant.

The through-line across the evidence is that partner discovery is rarely the bottleneck. Transfer architecture is. A firm can scan startups, run hackathons, and sign co-development MOUs while never resolving who owns what at the end. That resolution has to precede the first partner meeting, not follow it. You do not get to call yourself open because you bought a few founders coffee. You are open when the structure is clear enough that both sides are willing to share something that matters.

"We identify the problem that IBM was too closed in how they went to market. No software development kits, no APIs, no partners, no integrators."

— Henry Chesbrough, 20 Years of Open Innovation, CMR Channel, 2024 Chesbrough on IBM/OpenAI

IBM Watson is the canonical case of a company that built an ambitious natural-language AI system and then failed to extend it to external partners — no SDKs, no integrators, no IP-sharing framework that would let third parties build on it. Everything was done by IBM internally, and this limited greatly the markets the technology could address.

"The important lesson from looking at IBM and OpenAI is that it's important not just to be open with the technology itself, but in the marketing and commercialization of this and to bring on board others and other partners to help you explore several markets concurrently. That's going to give you a higher chance of success with any new technology."

— Henry Chesbrough, 20 Years of Open Innovation, CMR Channel, 2024 Chesbrough on IBM/OpenAI

IP governance is the mechanism that makes such commercialization extensions possible. Its absence is what capped IBM's addressable market.

How does open innovation differ from closed innovation — and what does open source have to do with it?

Open innovation is defined by purposively managed, selective knowledge flows. Closed innovation assumes that value is best created when all R&D stays within the firm's walls. Open source licenses knowledge to anyone without restriction. These are three distinct governance regimes, and mixing them up leads to the wrong strategy.

Source: Chesbrough (2003). Chesbrough & Appleyard (2007)

When does closed innovation still make sense?

Closed innovation still makes sense when disclosure risk is high, when interfaces are so tightly coupled that outside handoffs create more integration cost than learning value, or when regulation makes normal collaboration mechanisms impractical. The relevant question is never whether to be "open or closed" as a permanent philosophy. It is: which knowledge flows, in which direction, under which governance structure, for this product in this market?

Why open source is not open innovation

Firms that participate in open source (contributing to Linux, releasing code under Apache licenses) are not automatically practicing open innovation. West and Gallagher (2006) documented the paradox directly: firms that invest in open source R&D are funding outputs their competitors can freely use. Open source is a specific outbound mechanism where the firm gives up appropriability entirely. Open innovation selectively manages appropriability. The two are compatible strategies (many firms do both) but they are not synonymous.

What does inbound open innovation look like at scale? The P&G Connect+Develop case

P&G's Connect+Develop program is the most-cited example of inbound open innovation at scale and one of the clearest tests of whether the model can survive leadership change. By 2006, more than 35% of P&G's new products originated from external sources, and the program produced several billion-dollar brands before leadership transition stripped its infrastructure.

The problem Lafley faced in 2000

In 2000, A.G. Lafley took over as P&G's CEO. By then, internal R&D productivity had stagnated, and the pipeline that once generated brands like Tide, Crest, and Pampers was producing weaker returns despite continued investment. Lafley concluded that P&G could not build everything it needed inside its own labs, so the company made a structural shift and formalized external sourcing as a deliberate intake process.

How Connect+Develop worked

The program embedded technology entrepreneurs at P&G facilities to actively scout external innovations aligned with specific briefs. Unlike a passive idea box, C+D published what P&G was looking for and invited external partners to propose solutions. The commercial terms were explicit enough that P&G could license ideas in and take them to market quickly. The funnel behaved more like market validation than a suggestion box. Chesbrough later described Connect+Develop as the outside-in branch of open innovation.

The outcomes — and their limits

Key data block:

External sourcing produced named products such as the Swiffer Duster for cleaning tool design, Tide Cold Water through an external formulation partnership, Crest Whitestrips through licensed IP, and Pringles Prints through a printing-on-snacks technology licensed from a small Italian bakery. P&G brought Pringles Prints to market in under one year. Its typical internal timeline is two years.

"When I was writing this book that got published last year, I went back to P&G and the last 10 years have not been nearly as positive as the previous 10 years. So we rightly celebrated them in the 2000s but then we sort of lost interest or attention and in fact the last 10 years have been a real struggle for Procter and Gamble."

— Henry Chesbrough, Ep. 25 Open Innovation, WHU Podcast, 2021 Chesbrough on P&G decline

When Lafley's successor brought different priorities, the infrastructure that C+D required (leadership attention, partner management capacity, IP intake protocols) did not survive the transition. The lesson is not that open innovation does not work. Open innovation is an organizational capability, not a one-off program. Programs can be defunded. Capabilities decay when leadership stops tending them.

What does the outbound model actually earn? Tesla, Qualcomm, and the spectrum

Outbound open innovation ranges from deliberate ecosystem seeding to systematic IP monetization. Tesla and Qualcomm represent opposite ends of this spectrum, though Qualcomm's model sits at the extreme end where outbound shades into rent-seeking, making it a useful limit case rather than a canonical OI example.

Tesla's 2014 patent pledge

In June 2014, Elon Musk announced that Tesla would not enforce its patent portfolio against anyone who used its technology in good faith. The stated rationale was strategic, not altruistic:

"Our true competition is not the small trickle of non-Tesla electric cars being produced, but rather the enormous flood of gasoline cars pouring out of the world's factories every day."

— Elon Musk, Tesla Blog, June 12, 2014 Tesla blog (2014)

The pledge was not a license. It was a covenant not to sue, a narrower instrument that preserved Tesla's IP rights while removing the enforcement threat that would otherwise deter competitors from building on a shared charging and battery platform. Tesla blog (2014) Tesla's strategic calculation: growing the EV ecosystem grows Tesla's market faster than protecting IP from competitors who were, at that moment, largely irrelevant.

Qualcomm's licensing empire

Qualcomm's Technology Licensing (QTL) division generated $5.58 billion in revenue in FY2025 at a 72.4% earnings margin, roughly 25% of total company earnings from less than 13% of total revenue. Qualcomm FY2025 segment analysis Qualcomm's licensing revenue comes from standards-essential patents in mobile communications: chipmakers and device manufacturers must license Qualcomm's IP to produce devices that conform to 3G, 4G, and 5G standards.

This is active outbound licensing (IP as product) rather than ecosystem seeding. Qualcomm's model differs from Chesbrough's open innovation definition in a meaningful way: knowledge flows run in one direction and are enforced by patent obligation, not purposively managed for mutual benefit. It is a useful edge case precisely because it shows how outbound licensing can maximize value capture while contributing very little to shared learning.

Qualcomm's QTL licensing division generated $5.58 billion in FY2025 at a 72.4% earnings margin — roughly 25% of total company earnings from less than 13% of revenue.

The two models share the outbound label but pursue very different goals.

How does coupled open innovation work? IMEC, Pistoia Alliance, and Sanofi

Coupled open innovation requires a pre-competitive boundary decision (an explicit agreement on which knowledge is shared and which is withheld) before the first partner agreement is signed. The governance template is the program.

IMEC: semiconductor R&D at pre-competitive scale

IMEC is a Belgium-based research center that runs one of the world's largest coupled open innovation programs in semiconductors. Its annual budget is approximately $1 billion. More than 75% comes from 600+ industry partners, and its research staff exceeds 6,600 people from more than 100 countries. CSIS on IMEC

The model: member companies co-fund pre-competitive research (EUV lithography, process node development) that no single chipmaker could finance alone. Each member licenses the resulting IP under predetermined royalty structures. Competitive advantage is retained downstream, in proprietary chip design and manufacturing processes built on the shared pre-competitive base. IMEC has incubated approximately 300 startups and spun off more than 100. CSIS on IMEC

Pistoia Alliance: pharma pre-competitive data sharing

Founded in 2009 by AstraZeneca, GSK, Novartis, and Pfizer, the Pistoia Alliance now counts 200+ members, including 18 of the top 20 pharma companies by R&D spend. Pistoia Alliance Its purpose is to lower costs through pre-competitive collaboration on data and infrastructure that none of its members would share as a competitive differentiator.

Active projects include FAIR-principles standardization for in vitro pharmacology screening data (with FDA participation) and MethodDB for laboratory method standardization. The key insight is what is not shared: compound data, clinical trial results, proprietary efficacy findings. The pre-competitive boundary is explicit. Members share the infrastructure layer. They compete on everything above it. Pistoia Alliance

Sanofi iDEA-iTECH: IP option structure with academic partners

Sanofi's iDEA-iTECH Awards program in North America and Europe provides seed funding to academic partners and early-stage companies for breakthrough research. The IP structure follows a defined template. The academic partner retains initial IP ownership. Sanofi receives a right of first negotiation to license or acquire that IP when defined research milestones are reached. Sanofi iDEA-iTECH

This is coupled OI at the bilateral scale. Sanofi contributes capital and commercial development expertise. The academic partner contributes research capacity and IP. Risk and potential upside are shared, but IP assignment is predetermined. The governance template exists before the science begins.

What does the data actually show about open innovation performance?

The best-supported finding is that openness helps only to a point. After that, coordination costs and ownership ambiguity start to outweigh the gains. The numbers below show that tradeoff across adoption, completion rates, and value capture.

By the numbers

Open innovation performance figures from peer-reviewed sources:

The most important finding in this table is the inverted-U relationship from Laursen and Salter's 2006 analysis of UK manufacturing firms: searching widely and deeply is curvilinearly related to performance, and the relationship takes an inverted U-shape. Laursen & Salter (2006) Both insufficient and excessive openness underperform. The firms generating the best OI returns are those that tune the type and depth of external knowledge flow deliberately, rather than simply increasing it.

Laursen & Salter (2006) found that both breadth and depth of external search show an inverted-U relationship with innovation performance in UK manufacturing firms — beyond the optimal threshold, more openness reduces returns.

What kills open innovation programs? Five failure modes, quantified

The recurring pattern is not a shortage of ideas. It is unmanaged complexity. The failure modes below show what happens when firms widen their collaboration surface faster than they build the routines needed to absorb, govern, and measure what comes back.

Failure mode 1: NIH syndrome

Katz and Allen's 1982 study of 50 MIT R&D project groups is still the canonical quantitative evidence for the cultural resistance that kills inbound programs before IP governance even enters the picture:

"the tendency of a project group of stable composition to believe it possesses a monopoly of knowledge of its field, which leads it to reject new ideas from outsiders to the likely detriment of its performance"

— Katz & Allen, R&D Management, 1982 Katz & Allen (1982)

The performance pattern: team effectiveness increases through approximately 1.5 years of group tenure, remains roughly stable, then declines noticeably after the five-year mark. The mechanism is a reduction in communication with external sources of information. Long-tenured R&D groups stop paying attention to what is happening outside their walls, not because they are lazy, but because cohesion becomes insularity over time. Katz & Allen (1982)

Failure mode 2: absorptive capacity deficit

Cohen and Levinthal (1990) gave a compact definition of absorptive capacity: "a firm's ability to value, assimilate, and use new external knowledge." Cohen & Levinthal (1990) The catch is that firms cannot extract much value from outside knowledge unless they already possess related knowledge of their own. A firm that underinvests in internal R&D weakens its invention capacity and its ability to recognize and use valuable external inputs. Cohen & Levinthal (1990) This is the mechanism behind the common failure of firms that open inbound channels but cannot process what comes through them. Inbound OI requires a functioning receiver, not just an open door.

Failure mode 3: the inbound-only trap

Chesbrough identified the structural failure directly: for programs that successfully identify external opportunities, the hardest problem is not finding more ways to do open innovation. It is getting internal businesses to pay attention. Chesbrough on the inbound-only trap The obstacle is incentive structure, not communication. Business units have rational reasons to deprioritize externally sourced opportunities that add integration cost. Without routing authority, inbound OI produces a portfolio of interesting ideas that no one commercializes. Chesbrough on the inbound-only trap

Failure mode 4: partner incentive misalignment

In coupled and consortium arrangements, partner incentives diverge over time. When members contribute more to a consortium than they extract, they have rational incentives to reduce contribution or exit. The pre-competitive boundary that looks clear at launch often becomes contested as research outputs move closer to commercialization. Pistoia Alliance's longevity (17+ years) reflects disciplined maintenance of that boundary through shared infrastructure rather than competitive insights. Pistoia Alliance Consortia that drift into sharing competitive knowledge tend to fracture.

Failure mode 5: measurement vacuum

Programs that cannot demonstrate ROI cannot survive leadership transitions. Three standard KPIs for open innovation programs:

• External-source rate: what percentage of products or solutions involved external inputs
• Pipeline velocity: time from external identification to internal integration decision
• Time-to-integration: time from integration decision to commercialization

Programs that track only activity metrics (number of startup meetings, number of hackathon submissions) cannot answer the CFO's question. In MIT Sloan Management Review, Chesbrough makes the same point in plainer language: the hardest problems usually sit inside the firm rather than outside it. MIT Sloan Management Review Programs that cannot quantify their contribution do not survive the next cost-reduction cycle. That is where innovation feedback loops matter.

What are the most common misconceptions about open innovation?

The most persistent misconception about open innovation is equating openness with the absence of governance. Programs that produce measurable value are those with the clearest boundaries, not the loosest ones. Each misconception below has an empirical corrective.

Misconception 1: Open innovation means outsourcing your R&D

The belief: You can reduce internal R&D spend by sourcing innovation externally.

The correction: Cohen and Levinthal's absorptive capacity framework shows that external knowledge is only usable if internal knowledge exists. A firm that defunds internal R&D to fund external scouting destroys the receiver needed to process what external partners bring. Open innovation changes the sourcing mechanism, not the integration requirement. The firm still has to select, evaluate, and route what comes in.

Misconception 2: More openness is always better

The belief: A program looks better when it engages more partners, runs more hackathons, and collaborates with more universities than a more focused program does.

The correction: Laursen and Salter's inverted-U finding is the empirical refutation. Firms that search too widely dilute focus, increase coordination costs, and produce lower innovation performance than those that identify the right external search depth for their industry and capability level. Belgian CIS data from Van Criekingen et al. (2025) confirms that the firms with the most collaboration types have the worst project completion rates.

Misconception 3: Open innovation is the same as crowdsourcing

The belief: Running a public idea challenge is practicing open innovation.

The correction: Crowdsourcing is one inbound mechanism that may or may not constitute open innovation, depending on whether there is a governance structure for knowledge flows and IP assignment. LEGO Ideas is a crowdsourcing platform that operates as genuine inbound OI because it includes a defined IP structure (1% royalty to creators), a selection process, and a commercial pathway. A company is not practicing open innovation if it only runs an idea-submission form without an evaluation process, IP clarity, or a route from submission to commercialization. It is collecting suggestions.

Misconception 4: A high rejection rate means the program is failing

The belief: LEGO Ideas only launched 23 products out of 26,000+ submissions, so the program was badly designed.

The correction: 10,000 supporters is the filter. Before LEGO commits resources to formal review, the program asks whether an idea can attract visible community backing, and ideas that fail at that stage are ideas consumers have not endorsed. That is why the low selection rate is a design feature rather than evidence of dysfunction.

Misconception 5: Open innovation requires a dedicated innovation lab

The belief: You need a separate organizational unit (skunkworks, corporate innovation center, venture studio) to run open innovation.

The correction: Governance requirement, not facility requirement. P&G's Connect+Develop embedded technology scouts within the existing R&D organization instead of creating a parallel structure. IMEC is a standalone consortium, but member firms participate through their normal product development organizations. A dedicated lab is one option. It is not the prerequisite. The prerequisite is clear intake, IP assignment protocols, and internal routing inside existing structures.

When should you not use open innovation?

The decision to not pursue open innovation is as strategic as the decision to pursue it. Three conditions make open innovation a poor fit: when IP protection is weak, when internal R&D capacity cannot process external inputs, and when product architecture requires tight component integration that external sourcing would compromise.

Open innovation is likely the wrong choice if:

1. You cannot protect the value you disclose.
2. Your firm lacks the internal R&D capacity to absorb external knowledge.
3. Your product architecture requires tight integration between components that external sourcing would compromise.

When IP appropriability is weak

David Teece's 1986 appropriability framework remains the canonical analytical tool for this decision.

"innovating firms often fail to obtain significant economic returns from an innovation, while customers, imitators and other industry participants benefit"

— David J. Teece, Research Policy, 1986 Teece (1986)

When IP protection is weak, because patents can be designed around, because innovations are process-based and unpatentable, or because complementary assets are controlled by others, opening knowledge flows accelerates the transfer of value to competitors. Teece (1986) The Teece framework says: assess appropriability first. Design knowledge-flow architecture second.

When absorptive capacity is insufficient

Brunswicker and van de Vrande (2014) documented the specific constraints facing smaller firms in open innovation programs: "existing findings on large firms cannot be directly transferred to smaller enterprises." Brunswicker & van de Vrande (2014) Smaller firms face distinct absorptive capacity constraints, insufficient internal R&D, no dedicated partner management resources, and inadequate IP legal infrastructure. The practical implication: firms below a certain capability threshold cannot process external knowledge at the rate OI requires. For these firms, open innovation is not merely suboptimal. it generates coordination costs that exceed any knowledge benefit.

A study of 384 manufacturing SMEs across UAE emirates (Al Nuaimi et al., 2023) found that inbound OI positively impacted both market effectiveness and profitability, while outbound OI only affected profitability, not market effectiveness. Al Nuaimi et al. (2023) The asymmetry is a practical decision rule: for most firms below a certain size, selective inbound is the only viable starting point. Outbound and coupled programs need governance infrastructure first. Starting with multi-partner consortia before that infrastructure exists does not make a firm more innovative. It makes it more exposed.

When architecture requires tight integration

The edge case shows up most clearly in defense and national security contexts. Classification requirements, procurement law, and IP restrictions mean that neither inbound acquisition nor outbound knowledge flows can operate through the standard mechanisms OI programs use. BESA Center on defense OI Defense firms are not failing at open innovation. They are operating in a regime where the standard model is structurally constrained.

What comes next? AI as a knowledge broker in open innovation

The three structural variants (inbound, outbound, coupled) are not the final word on open innovation architecture. The newer literature treats AI as a broker layered onto the classic three variants: a way to search, match, and route knowledge faster across existing boundaries. Holgersson et al. (2024)

Holgersson, Dahlander, Chesbrough, and Bogers spelled out the mechanism in the companion paper from the same 2024 CMR issue:

"Artificial intelligence can enhance, enable, or replace traditional open innovation practices, changing the scope and efficiency of both outside-in and inside-out OI."

— Holgersson, Dahlander, Chesbrough & Bogers, California Management Review, 2024 Holgersson et al. (2024)

The structural difference from classic inbound OI is the reduction of human bandwidth as the bottleneck on search breadth. Holgersson et al. (2024) What AI-mediated knowledge flows do not resolve is the ownership question. The pre-competitive boundary decisions that govern IMEC and Pistoia consortia remain necessary. AI changes search costs faster than it changes ownership rules. Holgersson et al. (2024)

FAQ

For adjacent terms, browse the innovation concepts map, study the innovation flash cards, or reach for the business model canvas when the problem is venture design rather than knowledge flow.

What is open innovation?

Open innovation is about managed knowledge flows. Firms use it to move knowledge across organizational boundaries in ways intended to improve R&D outcomes, whether the flow comes inward from partners, universities, customers, and startups or outward through licensing, patent releases, and joint ventures. What defines the model is purposive governance of those flows, not just the fact of external collaboration.

What are the three types of open innovation?

The three structural variants are inbound (outside-in), outbound (inside-out), and coupled. Inbound means acquiring external knowledge. P&G's Connect+Develop is the canonical example. Outbound means releasing internal knowledge externally. Tesla's patent pledge illustrates ecosystem-seeding outbound, while Qualcomm's licensing division shows IP-as-product outbound at its most extractive. Coupled means bidirectional flows managed through alliances or consortia. IMEC and the Pistoia Alliance are canonical coupled programs.

How is open innovation different from a standard R&D partnership?

An R&D partnership is a bilateral agreement between two firms to co-develop a specific product or technology. Open innovation is a broader organizational model that manages multiple inbound and outbound knowledge flows at once, with explicit IP governance structures for each direction. An R&D partnership can be one component of an open innovation program. It is not equivalent to the program itself.

What are the risks of open innovation?

The primary risks fall into three areas. One is IP leakage, where internal knowledge flows outward through inbound channels. Another is absorptive capacity deficit, meaning the firm lacks the internal R&D capability to process external inputs. A third is the inbound-only trap, where the front-end scouting process finds opportunities that the internal organization is not equipped or incentivized to act on. A 2025 Journal of Technology Transfer study found that Belgian firms using the most collaborator types had project completion rates of 44% versus 60% for firms collaborating with no external partner types, a reminder that more openness without more structure can produce worse outcomes, not better ones. Van Criekingen, Freel & Czarnitzki (2025)

How do you protect IP in an open innovation program?

IP governance structures differ by variant. Inbound programs require NDAs before disclosure and IP assignment clauses defining what happens to derivative works. Outbound programs require explicit licensing frameworks specifying what is licensed, to whom, on what terms, and whether the license is exclusive. Coupled programs require pre-competitive boundary agreements drafted before any joint research begins, specifying which outputs are shared and which remain proprietary. Those rules should exist before the first partner conversation starts. Teece (1986)

Is open innovation right for smaller companies?

With significant caveats. Brunswicker and van de Vrande (2014) found that large-firm OI frameworks do not transfer directly to SMEs, which face distinct absorptive capacity constraints, insufficient partner management resources, and inadequate IP infrastructure. Brunswicker & van de Vrande (2014) Selective inbound programs (engaging one or two external partners in structured pilots with explicit IP terms) are more tractable for smaller firms than multi-partner consortia. Coupled and consortium models require organizational infrastructure that many firms below 500 employees do not yet have.

How do you measure whether an open innovation program is working?

Standard metrics include external-source rate, which measures the percentage of products or solutions with external-origin components, pipeline velocity, which measures the time from external identification to an internal integration decision, and time-to-integration, which measures the time from that decision to commercialization. Programs that track only activity metrics, such as the number of partnerships or submissions to a challenge, cannot demonstrate value to the organization. ROI requires comparing the cost of external knowledge with the cost of developing that knowledge internally, adjusted for time-to-market and quality.

Colaboradora

Clara @cla_reinholt

Se centra en la comunicación de innovación, la facilitación y en convertir los marcos en hábitos de equipo.

Clara writes about the human systems behind innovation: facilitation quality, communication clarity, and the routines that help teams move from ideas to decisions. She follows practical team-method sources such as the Atlassian Team Playbook, alongside innovation coverage from McKinsey and Harvard Business Review.

Her contributions often combine editorial storytelling with practical templates that leaders can reuse for team rituals, retrospectives, and portfolio reviews, informed by research and practices from McKinsey on Innovation, Harvard Business Review, and the Atlassian Team Playbook.

Clara tends to ask one recurring question in her drafts: Will this help someone lead a better conversation tomorrow? If the answer is yes, the piece is ready.