#95 Models for AI robotics adoption

00:00:03: Welcome to IP Management Voice, the podcast spotlighting leading minds in IP and IP management.

00:00:09: This series supports The IP Business Academy's call for subject matter experts To share insights foster reflection And strengthen a global IP system.

00:00:19: Join the conversation via our newsletter and explore cutting-edge content on Deplex.

00:00:24: this episode which explores how companies can decide whether to develop co-develop or purchase knowledge For AI robotics adoption features content from David Aldrich.

00:00:36: If you are managing an IP portfolio or directing a R&D division, running tech due diligence right now... Your default instinct is probably to build a fortress.

00:00:46: Right,

00:00:47: absolutely!

00:00:47: That's the standard playbook.

00:00:48: Yeah

00:00:49: exactly The Standard Engineering and Legal Mindset basically dictates that you identify a core technology You draft the widest possible claim scope Secure your freedom to operate And just aggressively guard your trade secrets Which

00:01:02: makes total sense.

00:01:02: in a vacuum

00:01:03: It does.

00:01:04: But um... A massive new study reveals when it comes to adopting AI robotics that traditional Fortress mentality Is actively suffocating innovation

00:01:13: working against you.

00:01:13: Exactly, so today we are doing a deep dive into a really compelling twenty-twenty six paper published in technological forecasting and social change is by odrish blitzky and rigid

00:01:25: right titled unlocking AI robotics.

00:01:27: adoption in innovative firms.

00:01:29: make buy or ally yes

00:01:31: And the data set here just demands serious attention.

00:01:34: We're talking about micro level data from over fifteen thousand Innovative UK SMEs

00:01:39: yet analyzed over a sixteen year period From two thousand four to twenty.

00:01:43: That is huge.

00:01:43: Okay, let's unpack this because the scale of that data set is just

00:01:47: wild.

00:01:48: The longitudinal nature Of that data Is exactly what makes This so compelling for strategic decision makers.

00:01:54: I mean we are looking at a handful of isolated case studies with survivorship bias.

00:01:59: Right, it's not just a couple of Silicon Valley unicorns

00:02:02: exactly.

00:02:02: we are looking at a macro level shift in how firms actually capture value when dealing with general purpose platform technologies

00:02:10: and the findings directly challenge How We Traditionally Value Patent Portfolios And well structure technology life cycles entirely.

00:02:18: they really do.

00:02:19: The stakes for your patent portfolio are incredibly high here.

00:02:22: So our mission for this deep dive is to completely dismantle the assumption that highly innovative firms have to rely on pure internal development, to maintain their competitive edge?

00:02:32: Because that rigid IP lockdown mentality is actually constraining them

00:02:35: Exactly.

00:02:36: We're going explore why purely internal R&D strategies Are a bottleneck For AI robotics And how strategic tech transfer specifically through external acquisitions and active alliances is actually winning the race.

00:02:49: To understand that shift, I think we have to look really closely at the mechanics of why that traditional foundation is cracking under the weight of AI robotics.

00:02:58: complexity?

00:02:59: Yes!

00:02:59: So let's start with that traditional Foundation The MAKE strategy...the purely internal R&D engine.

00:03:07: In a classical innovation model, Internal R & D investment is the undisputed driver.

00:03:14: You fund the lab, they build a tech you deploy it.

00:03:16: That is the expectation.

00:03:18: but The paper's analysis of internal R&D investment reveals something highly counterintuitive Once you factor in external acquisition and collaboration channels Internal r&d does not independently raise the odds of AI robotics adoption.

00:03:32: Wait!

00:03:32: Not at all?

00:03:33: Not independently.

00:03:34: No It Is Wild

00:03:35: that completely forces A total reevaluation Of how innovation managers allocate their budgets.

00:03:40: I mean, we typically view internal R&D as the spark that drives implementation.

00:03:45: But for complex adaptable technologies like AI Robotics it functions primarily as a baseline capability.

00:03:52: In literature this actually aligns with Cohen and Leventhal's concept of absorptive capacity.

00:03:57: Oh, absorptive capacity.

00:03:59: Right remind me how that applies here.

00:04:01: what

00:04:01: you basically need a baseline of internal technical knowledge just to recognize the value of an external technology.

00:04:09: You have to be able to assimilate it into your own operations

00:04:11: I see.

00:04:12: so you need engineers Just understand What are buying

00:04:15: exactly.

00:04:16: Pouring capital into internal R&D alone is not the mechanism that actually scales the robotics on your factory floor.

00:04:22: It just prepares you to absorb

00:04:24: it.

00:04:24: Okay, let me push back a little bit or at least look from the perspective of an R and D director.

00:04:29: Sure If I am looking my staffing And my internal R & D isn't primary engine for adoption.

00:04:35: Who was doing heavy lifting?

00:04:38: The demographic data in this study has a really fascinating answer to that.

00:04:42: Yeah, the breakdown between STEM and non-STEM work?

00:04:44: Exactly!

00:04:45: It shows that while STEM workers... you know your data scientists and algorithmic engineers are a baseline requirement ...the share of non-stem workers is actually more consistent predictor.

00:04:56: Wait,

00:04:56: really?!

00:04:57: Non-STem workers are better predictors for successful AI robotics adoption?

00:05:01: Yes it's much more robust predictor.

00:05:04: And that data point gets to the absolute core of the implementation friction.

00:05:09: Because

00:05:09: the bottleneck isn't the algorithm itself,

00:05:11: right?

00:05:12: The bottleneck in AI robotics adoption is rarely the fundamental math.

00:05:17: The friction is entirely in the integration

00:05:19: process redesign change management That sort of thing precisely

00:05:23: figuring out how to integrate new autonomous routines.

00:05:27: It's a highly rigid legacy workflows.

00:05:29: The non-STEM workforce represents the managerial and operational capability required to translate a theoretical model into a functioning service pipeline.

00:05:38: Oh, that makes total sense because if your warehouse managers and floor operators reject the automated routing system... ...because it doesn't align with their daily constraints?

00:05:45: The most elegant STEM engineering in the world won't save the deployment.

00:05:49: It will just sit there.

00:05:50: Sounds like internal R&D is no longer the underlying software engine.

00:05:53: I kind of view it as an API.

00:05:57: That's great analogy.

00:05:59: You need it so the external tech doesn't crash your legacy system, but its not core product anymore.

00:06:05: You aren't building a smartphone you are just downloading the app.

00:06:08: Exactly!

00:06:09: Your internal R&D is integration architecture.

00:06:11: But wait if I am an investor evaluating a firm's IP and their internal R & D is operating merely as an integration layer haven't they effectively hollowed out there own technological core?

00:06:23: It looks like that from a traditional perspective, yes.

00:06:25: Right because where is their competitive moat if they aren't generating pure foundational inventorship?

00:06:31: What's fascinating here is that we have to fundamentally redefine where the competitive moat actually resides for general purpose technologies.

00:06:39: Okay,

00:06:39: how so?

00:06:40: Well AI robotics is highly adaptable across multiple sectors which creates massive inherent appropriate ability challenges

00:06:47: because The technology frontier diffuses and evolves So rapidly.

00:06:51: right.

00:06:51: if a firm pours its capital into purely internal make strategies They're absorbing.

00:06:56: they total crushing cost of specialized AI personnel Plus, massive data infrastructure and constant algorithmic experimentation.

00:07:05: And what happens practically?

00:07:06: What happens is they end up with brilliant heavily patented prototypes that never actually scale simply because the broader technology landscape moves way faster than their isolated internal team can iterate.

00:07:19: So, they essentially exhaust their operational budget trying to maintain sole inventorship over a system that is obsolete by the time the patent has even granted.

00:07:27: Exactly!

00:07:28: The competitive moat is no longer holding a foundational algorithm patented in a vault.

00:07:33: Wow… then what's it?

00:07:34: The true moat was the firm's dynamic capability—their absorptive capacity —to successfully integrate external knowledge faster and more effectively than competition.

00:07:45: Purely internal R&D won't scale AI robotics.

00:07:48: The only logical pivot is to look outward and leverage the broader ecosystem,

00:07:52: right?

00:07:52: You have to look outside of the fortress

00:07:54: but opening the gates in looking outward introduces massive operational risk.

00:07:58: The paper breaks the external approach into two distinct strategies by an ally.

00:08:03: let's Look at the buy strategy first because the data here Is the most unambiguous finding in the entire study.

00:08:09: it really is Purchasing.

00:08:11: external R&D is the absolute most consistent statistical driver for AI robotics adoption.

00:08:17: It acts as an absolute baseline accelerator, whether that means acquiring external tech outright in licensing intellectual property or utilizing specialized subcontractors

00:08:28: mechanically.

00:08:29: buying your way in just completely bypasses the talent war, doesn't it?

00:08:33: Totally.

00:08:33: You aren't trying to outbid major tech conglomerates for a handful of specialized machine learning engineers.

00:08:38: right you are purchasing a validated commercially tested solution and then you deploy your internal AP your R&D team strictly To integrate that purchase solution into your specific data architecture.

00:08:51: It

00:08:51: functions as an immediate high-fidelity learning mechanism because external providers bring the cumulative experience of prior deployments across dozens of different environments.

00:09:00: Which

00:09:01: allows SME to translate broad theoretical AI possibilities into highly specific operational tasks.

00:09:07: Exactly, completely avoiding massive adjustment costs and all those dead ends associated with exploratory ground-up R&D.

00:09:13: But buying an external solution means you have integrate black box third party technology With your proprietary data

00:09:20: And that instantly creates friction.

00:09:22: Right, which pushes firms toward the ally.

00:09:25: strategy knowledge collaboration and co-development.

00:09:27: And here's where it gets really interesting!

00:09:30: The

00:09:30: Ally Strategy has a massive caveat.

00:09:32: Yes...the

00:09:33: paper reveals a highly nuanced actually dangerous trap regarding how firms absorb external information.

00:09:41: The data draws a severe distinction between active collaboration and passive collaboration.

00:09:46: Right, so active collaborations like co-designing a specific use case with the key customer significantly boost adoption.

00:09:53: but passive knowledge spillovers actually have a negative coefficient.

00:09:57: They actively reduce the likelihood of adoption

00:09:59: In tech due diligence.

00:10:01: we actively look for that.

00:10:02: what I call innovation tourism.

00:10:04: We map accompanies technology landscaping in their conference presence to gauge their technological readiness.

00:10:09: Sure

00:10:10: ambient is traditionally seen as a good thing.

00:10:12: But this data set proves that simply monitoring the landscape, it's practically toxic for AI adoption.

00:10:18: Why does passive knowledge spill over backfire?

00:10:20: If we connect us to the bigger picture We really need to define those passive spillovers mechanically To understand their risk.

00:10:27: Okay

00:10:27: let bring you down.

00:10:28: In traditional linear innovation models, ambient learning is universally beneficial.

00:10:33: Sending your engineers to technical conferences monitoring trade standards reading scientific journals

00:10:39: keeping a pulse on the industry right.

00:10:41: but for complex platform technologies like AI robotics it comes down to deep operational entanglement.

00:10:49: this Is not a discreet plug-and-play component Like A new microchip.

00:10:53: ah Right It has To integrate With everything

00:10:56: Exactly.

00:10:57: AI robotics must be deeply integrated with a firm's specific data architecture, its privacy compliance and it physical operational routines.

00:11:05: So ambient spillovers just create overwhelming noise?

00:11:09: Yes A team attends an AI summit They observe a dozen different deployment architectures across various industries And they bring that unstructured knowledge back to the firm

00:11:17: Which creates profound misalignment in analysis.

00:11:20: paralysis Because you try to reverse engineer a competitor's architecture based on the fifteen-minute conference presentation, but lack their specific proprietary training data.

00:11:31: And integration inevitably fails!

00:11:33: The project stalls for eighteen months and your adoption odds just plummet...

00:11:38: It is attempting copy output without understanding underlying environmental fit.

00:11:43: Unstructured knowledge gathering with anchor of a specific active partnership generates mixed signals.

00:11:50: It causes managers to hesitate or misdirect their capital.

00:11:54: So the strategic mandate here has to shift from ambient learning, to selective capture?

00:11:59: Yes exactly!

00:12:00: Let's define selective capture practically because if I am an R&D director how do i actually filter that noise?

00:12:05: it requires implementing strict kill-or scale gates within your innovation pipeline.

00:12:09: okay what does that look like in practice?

00:12:11: Instead of a broad mandate to monitor AI trends, an SME takes ambient knowledge and immediately forces it through structured active collaboration.

00:12:20: Like a pilot program?

00:12:21: Yes!

00:12:22: like running tightly governed ninety day pilot programs with specific external supplier or co-designing singular narrow use case with an anchor customer.

00:12:32: And if the pilot doesn't hit predefined integration metrics,

00:12:35: you kill it immediately.

00:12:36: active structured alliances provide them necessary reality check.

00:12:40: passive information gathering just consumes resources and creates confusion.

00:12:45: that

00:12:45: operational reality check makes total sense.

00:12:48: but if active external collaboration deeply co-developing with suppliers and customers is the true key to adoption.

00:12:55: This where it gets tough for the legal team?

00:12:57: Yeah, this has to be triggering major alarm bells For any IP professional listening right now.

00:13:02: Oh absolutely

00:13:03: because If I'm an IT manager my entire compensation structure And My core KPIs are based on securing broad claim scopes maintaining absolute freedom To operate and preventing Any leakage of our trade secrets.

00:13:14: Right your job Is to build the fortress.

00:13:16: exactly When you tell my R&D team to aggressively co-develop, I'm terrified of losing pure inventorship.

00:13:22: This introduces what the paper identifies

00:13:33: as You'll be supported in creating digital content, LinkedIn strategies and outreach to your target group.

00:13:53: If you're ready to raise your profile and position yourself as a thought leader get in touch with us.

00:13:59: We'd love to hear from you.

00:14:01: End of advertisement.

00:14:05: This raises an important question And it is a statistical sign flip that fundamentally alters how we must structure IP strategy within collaborative ecosystems.

00:14:15: Let's look at the mechanics Because it is wild.

00:14:20: The paper demonstrates that when a firm is acting entirely alone, relying strictly on the internal need make strategy strong.

00:14:27: appropriability actually has positive association with adoption.

00:14:31: Right If you are operating in an isolated silo strict patenting and tight non-disclosure agreements help implement technology.

00:14:39: The logic there is sound within an isolated framework, I guess.

00:14:42: It is because if a firm is bearing the total unmitigated capital cost of internal R&D they require absolute assurance that they'd protect future returns on this massive investment.

00:14:52: Sure!

00:14:52: Strong rigid IP protection provides necessary confidence for board to actually deploy capital into costly adoption cycle?

00:14:59: Exactly.

00:15:00: But when you factor active knowledge collaborations with ally strategy in statistical model association entirely reverses

00:15:07: The sign flips.

00:15:09: Once a firm starts relying on external partnerships, strong appropriability actually reduces the odds of AI robotics adoption.

00:15:16: The core mechanisms designed to protect the firms' assets become the exact friction points preventing implementation.

00:15:23: It is critical vulnerability!

00:15:26: The rigid IP frameworks that managers rely on are actively hindering the firm's operational agility.

00:15:32: But wait let me challenge practical application.

00:15:35: We can't seriously be suggesting that IP directors should just open source their tech and abandon asset protection altogether.

00:15:42: Oh, definitely not!

00:15:43: Because no rational board of directors is going to approve a strategic alliance if it means exposing their core proprietary data to third-party vendor without aggressive legal safeguards

00:15:53: Of course not.

00:15:54: So how do we reconcile the fundamental fiduciary duty to protect shareholder value with this data?

00:16:00: showing strong IP protection kills AI adoption?

00:16:03: Well, the data isn't suggesting the abandonment of intellectual property.

00:16:07: It is exposing the flaw in a monolithic lockdown IP mindset.

00:16:12: Okay

00:16:12: so it's about how the IP has

00:16:13: structured Exactly.

00:16:15: When an IP department default setting Is demanding overly broad claim scopes Enforcing aggressive non-disclosure paranoia And refusing to engage in joint inventorship

00:16:26: they fundamentally break the necessary co-creation process.

00:16:30: Right, it is about shifting to collaboration compatible IP.

00:16:35: So we're talking about the friction of a contracting phase?

00:16:38: Yes!

00:16:38: The negotiation phase where these projects go die

00:16:41: For our non lawyer listeners.

00:16:43: when you enter joint development agreement background IP is pre existing technology.

00:16:48: bring to table and foreground IP what you invent together during alliance.

00:16:53: If an IP department spends eight months locked in a legal battle demanding exclusive ownership of all foreground IP, the external vendor will simply restrict access to their core API.

00:17:04: Exactly and The Legal Cycle outpaces the technology cycle.

00:17:07: Oh

00:17:07: that's great point!

00:17:08: By the time lawyers finally agree on the ownership structure... ...the specific AI model the firm was attempting implement is functionally obsolete.

00:17:16: Wow

00:17:16: so A rigid ip portfolio that suffocates the integration process Is actually massive liability.

00:17:22: It IS.

00:17:23: firms must utilize modular intellectual property architectures instead.

00:17:28: Mechanically, what does a modular IP architecture look like in an AI alliance?

00:17:33: You establish absolute clean boundaries around your core proprietary training data that remains closely guarded.

00:17:40: trade secret always

00:17:41: right.

00:17:42: you protect the ground jewels

00:17:43: exactly but you'd utilize stage disclosures Open join IP clauses and flexible data governance agreements for the integration layers.

00:17:51: So you are more flexible on the edges?

00:17:53: Yes,

00:17:54: it involves utilizing cross licensing to ensure mutual freedom To operate specifically within the confines of The Alliance

00:18:01: rather than attempting to assert global vertical dominance over the entire technological application.

00:18:08: Right!

00:18:08: You don't need to own everything everywhere just what you need to operate the robotics in your specific context.

00:18:14: It's basically setting up a data governance structure that allows an external partner to train their machine learning model on your proprietary data without that partner actually gaining ownership of the underlying dataset itself.

00:18:26: Precisely, The goal of the IP strategy must shift from pure exclusion To facilitating rapid secure iteration with external partners.

00:18:37: Because the data proves that the cost of extreme appropriability in a networked environment is a significantly lower likelihood of actually getting robotics to function on your floor!

00:18:47: If ultimate goal is adoption and commercial scaling, The legal framework absolutely has to serve operational reality not hinder it.

00:18:55: So a firm's capacity to navigate joint inventorship fluidly is now more critical asset than static patent vault.

00:19:02: I would argue yes, absolutely.

00:19:04: That requires complete paradigm shift for R&D and IP leadership.

00:19:08: Let's synthesize the core mechanics of this deep dive because it fundamentally rewrites standard playbook.

00:19:13: It

00:19:13: really does turn conventional wisdom upside down.

00:19:15: In the realm of highly complex platform technologies like AI Robotics, your purely internal R&D is no longer your adoption engine.

00:19:22: It's your baseline integration readiness...

00:19:25: ...it's the API that allows YOUR firm to absorb external innovation without catastrophic failure!

00:19:30: Right

00:19:30: and you check up your strategic ability to acquire validated external R & D as your primary accelerator bypassing the talent war entirely.

00:19:39: And your active alliances specifically deep co-development with supply chain partners are your operational reality check,

00:19:46: actively filtering out the paralyzing noise of passive ambient knowledge spellovers and

00:19:51: tying all those elements together.

00:19:53: A rigid highly secretive IP strategy will only serve to isolate you firm.

00:19:59: Attempting to maintain sole inventorship and maximum claim scope over an integrating technology will just cut you off from the collaborative ecosystem you desperately need to actually scale.

00:20:09: It is a recipe for being left behind,

00:20:11: which leaves us with a critical strategic question to ponder especially if you are heading into M&A negotiation or boardroom strategy session this week.

00:20:20: If the data conclusively shows that most successful highly innovative firms are moving away from heavily protected, purely internal development and shifting aggressively toward collaborative modular technology integration.

00:20:32: How does this alter the fundamental mechanics of IP due diligence?

00:20:36: Exactly.

00:20:37: In the next five years, will we continue to value an enterprise based on the sheer volume of patents they hold defensively in a vault?

00:20:44: Or

00:20:44: will market valuations strictly reflect their collaborative integration capacity and the operational fluidity?

00:21:07: Please get in contact and let us know.