Google vs Microsoft
Google and Microsoft have taken two different paths to commercialize generative AI technology, and we are currently witnessing a real-life case study of their approaches.
Google, which had been the front-runner in the AI space since its 2018 announcement about LaMDA, chose to incubate and build the technology in-house. But their recent responses to ChatGPT have shown signs of panic, with a Code Red in January and a chaotic Bard demo event in February.
Meanwhile, Microsoft ‘invested’ $1 billion and purchased the exclusive right to OpenAI’s underlying technology in 2020. A subsequent $13 billion check was reported to give Microsoft 75% of OpenAI’s profits until it secures a return on its investment and a 49% stake in the capped-profit company. These deals are already sparking conversations about how the acquisition of OpenAI by Microsoft is almost inevitable.
Considering just that part of the story, it’s a tale of how two powerful incumbents adopted different corporate venturing strategies: building innovation in-house versus acquiring it externally (through partnerships, equity investments and/or acquisitions); and in the media frenzy that has followed Microsoft’s $13 billion investment, much has been written about how Microsoft has won the generative AI race and how Google’s innovator’s dilemma led it to fall behind.
What gets lost in that part of the story, however, is that Microsoft tried building generative AI innovations in-house first, but failed quite miserably. It fell into the same pitfalls that many corporations fall into when building innovation internally- leaving it with little choice but to form an alliance with an external startup to catch up with other incumbents in the space.
Microsoft’s foray into generative AI started in the early 2010s, with its efforts dispersed amongst different teams before being centralized, in 2016, under the Microsoft AI & research group, which was staffed with some 5,000 engineers and grew to 8,000 a year later. Multiple attempts were made to build generative AI products internally, including Tay (2016), Zo (2016), and Rinna (2015-2018); each of those ended up being shut down for similar issues – the chatbot coming unhinged.
While Google also had its own debacles (Google Duplex, for one), it eventually changed tack and started infusing generative AI a lot less explicitly into its existing products, like Gmail (composing emails, for example) and Google photos (for organizing pictures). These product improvements indicate that Google had been years ahead of its competitors.
Launching new businesses takes continual commitment; it's not an activity for the impatient or for the fainthearted
- Ralph Biggadike (Harvard Business Review, 1979)
What were the pitfalls that Microsoft (and many other organizations) fall into when pursuing new technologies and ventures internally? and how can they be avoided?
To answer these questions, we spoke to innovation leaders in 11 different corporations across industries ranging from agriculture to aviation, and energy to industrial materials. Each of these companies has a slightly different approach to internal innovation but, perhaps unsurprisingly, they all seem to face similar challenges.
We focused our conversations on internal corporate venture (ICV) units that have been set up separately from the core business. These ICVs incubate ideas that originate from within the organization (and are, at the very least, adjacent to the core) but sometimes threaten to disrupt the core.
We looked into the following aspects of ICV efforts:
- People: How companies staffed, supported and incentivized talent within these ICVs to take on the challenges of setting up a new ‘venture.’
- Process: How they shepherded ideas through the different stages of venture development – from validating early ideas and exploring technologies to actual commercialization.
- Governance: How they made decisions and funded internal ventures.
To get a full overview of the details of the approach taken by these companies, download the report here. (The report has been anonymized so as to protect confidential information.)
Here’s a summary of the key challenges we found, along with the tried and tested approaches to overcome them:
How to avoid unclear direction and strategy
Vacillating between the centralized and dispersed organization of internal innovation is nothing new; many of the organizations we spoke to have done the same. But the ones that established an organizational format that succeeded were the ones that looked strategically at what they wanted to achieve.
Through our experience in helping multiple organizations define their innovation strategies and objectives, we offer the following rules of thumb: Improving your core offerings falls neatly within the purview of current business strategy. Reviving a lagging product or a product portfolio, or continually launching innovative products and services to renew your core product portfolio – these are often the core concerns of business leaders and typically enjoy the support of the ‘mothership’. So if your main objective is to improve your core, then the best organizational approach is probably an internal innovation team, fully integrated within existing business units, with a clear mandate and process, and with senior-level leadership involvement.
But disrupting the core or finding breakthrough adjacent business won’t enjoy the same internal support; therefore making it the responsibility of the core business leaders is a recipe for disaster. So if your main objective is to find new growth areas adjacent to the core, explore innovations that fit with the long-term vision (>7-10 yrs) of your organization (but not necessarily with the short-term (<5yrs)), or that disrupt your current core business entirely). You should centralize your innovation efforts by setting up an external ICV unit, separated from the core business units in both processes and funding. You can leverage the resources of the parent organization (funding, talent, expertise) by hiring some ICV team members from the core organization and having executive-level leadership of the ICV unit. This approach can also be used for ‘closer-to-core’ innovations which involve extremely new or unknown technology.
Existing businesses can't take [in] these ideas when they 'graduate' because they just don't have the right leadership, structure, or capabilities. It gets squeezed out by the core.
- Venture Leader at a new digital ICV Studio
How to avoid undue pressure to ‘integrate’ (or spin back in):
Even when organizations set-up an external ICV unit to explore adjacent businesses, there is often undue focus on eventual re-integration into the core business.
When speaking to executive leaders, we often notice a strong resistance to spinning out. After all, the internal venture represents an investment of resources and talent, and spinning it out will mean its success is not reflected on the balance sheet of the mothership.
However, experience teaches us that ICVs, when spun into existing business groups, are often regarded as resource black holes until they generate profits; and/or as disturbances to the core business and processes; and/or as foreign ideas/teams that weren’t ‘invented here’ and therefore are sub-par to allocate resources or attention. This leads to a corporate ‘organ rejection’ of the new ventures.
You need to buy-in from the very top leadership. As you navigate outside core business, everything below that... is enemy country.
- Head of ICV arm
The choice, between spinning in and spinning out, in the minds of executive leadership seems to be between ‘letting go’ of invested resources or keeping them in-house. In our experience, the choice is between ‘letting go’ of ventures to give them a fighting chance at survival or keeping them in-house and drastically reducing their chance of success.
Unless your industry is faced with a technology that poses the possibility of a real paradigm shift, as AI did for Microsoft and Google – the decision to spin-in or spin-out may be well outside of the purview of ICV units. So how can you spin in an ICV without almost guaranteeing its failure?
We suggest two approaches, inspired by the leading companies with whom we spoke:
- Hold BU leaders individually accountable for the transition plan. The ICV arm of a large chemicals player we spoke with establishes a formal MoU and puts expectations from the business unit in writing. This ‘forces’ BU leads to honor their responsibility towards nurturing an ICV when spun in.
- Introduce ‘New business development’ KPIs in the accountabilities of the BU leaders. For instance, in the case of a large materials manufacturing company, Business Unit leads were assigned performance targets so that a significant percentage (30%) of their revenue would come from products or services launched in the last four years. These performance metrics can incentivize leaders to support new ventures instead of ignoring or trying to kill them.
These two approaches may be ‘politically charged’, prompting BU leaders to resist, since ambidexterity can be an uncomfortable skill to acquire; but this is a small price to pay when the alternative is a leadership organization that is stubbornly committed to the status quo when your organization needs change.
How to avoid restrictive internal policies
Internal corporate policies are built to minimize risk and increase efficiencies for the core business; invariably, they end up slowing speed to market.
Two research engineers at Google developed (and wanted to launch) a ChatGPT-like product five years ago. Their plans were thwarted because their product did not adhere to the company’s AI safety standards; this institutional resistance was probably l sensible given that a factual error at the BARD announcement cost Google 100 billion in market value. However, in its bid to respond quickly to Microsoft, it may have swung too far to the opposite extreme – reports indicate that, this January, Google’s ‘Code Red’ prompted a change in internal policies to introduce a “green lane” to expedite the approval process for AI-related products including assessing and mitigating potential harms.
Some internal policies around ethics and reputation risk management are relevant- not only for the corporate mothership and its internal ventures but also for external startups, especially with respect to a technology that poses as many risks as AI does to society. Other internal rules, such as restrictive procurement, hiring, and legal policies, might not be as essential.
The most eloquent description of this is from Noam Bardin’s (ex-CEO Waze) article on why he left Google:
The amount of time and effort spent on Legal, Policy, Privacy - on features that have not shipped to users yet, meant a significant waste of resources and focus. I am not saying this is not important, but this had zero value to our users. An ever-increasing percent of our time went to nonuser value creation tasks, and that changes the DNA of the company quickly, from customer-focused to corporate guidelines focused.”
So how do you strike the right balance between risk mitigation and upside potential?
Here are a few approaches we have heard about from our interviewees and helped implement at our clients’ organizations:
- Autonomy from internal processes: Creating a wholly owned subsidiary can provide the link to the mothership that the ICV needs while allowing sufficient independence from corporate processes to allow for speed.
- Systematize exceptions to internal policies or ‘Green Lanes’: It might be a good idea to think of a ‘Green Lane’ before your biggest competitor tries to steal the carpet from under your feet. With enough ICV teams having gone through the process, you can identify the critical junctures where internal processes do not work. Engage with the functional leadership and teams to ‘systematize’ exceptions, creating a standard operating procedure for ICV units and how (and how far) exceptions will be created.
How to integrate the venture with the internal processes from HR to finance, to legal? We have this saying, if it walks like a duck and it quacks like a duck, it's a duck. If you want it to be a start-up, let it be a startup.
- VP of Innovation and Ventures
How to avoid insufficient resourcing
- Part-time innovators: Far too many organizations expect employees dedicating only 20-30-50% of their time to ICVs to be successful. Even more surprising, often no effort is made to reduce the workload from their ‘day job’. This inadvertently means the employee needs to work on the ICV ‘on top of’ her day job – a situation that is untenable for a short period, let alone for the length of time it takes to build a venture.
- Incentives: Many corporations assume that employees are given a riskless opportunity when they join an ICV – the ability to start their own startup within the mothership while having job security. This is not entirely true. Every ICV employee risks their otherwise stable corporate career – because even though they have job security they don’t have promotion security. As one of our interviewees put it:
[By joining an ICV] they don't climb up the corporate ladder. They have done something unconventional and [if] it failed, the label of failure will be on your back.
- Head of ICV arm
- External support: Paradoxically, many companies place existing leaders as mentors in their incubation and acceleration programs -the very people who have built their entire careers in corporate life. Mentorship from real entrepreneurs, who have gone through the actual grind, was something that only a few of the organizations we spoke to actually provided.
The suggestions to avoid this pitfall are difficult to execute, but it’s important to remember that breakthrough ideas are never easy to execute.
- Mix internal with external talent: We heard many interviewees talk about the lack of availability of entrepreneurial talent within their organization. While this may well be used as an excuse to avoid doing any impactful innovation, if this is indeed the case, external recruitment is always an option. The trick is to recruit your new employees for their entrepreneurial skills, not industry experience. The new recruit can bring the right skills and entrepreneurial experience and you can supplement their skills with your best internal talent to bring in domain and organizational knowledge.
- Incentivize appropriately: If you recruit entrepreneurs, internally or externally, you need to incentivize them like entrepreneurs. This does not mean 100% equity ownership in the new venture, but it does mean some participation in success (eg. through a percentage of the contribution margin).
- Guide adequately: Build a database of external, entrepreneurial mentors as an essential feature of your ICV unit. This is a need expressed by most ICV teams but typically is only provided by external (non-corporate) venture studios.
- Provide full-time resources: Bringing team members from the mothership part-time is effectively setting up the team for frustration and, eventually failure.
How to overcome the mid-level leadership challenge
The average ICV takes ten years to deliver significant ROI. BU leadership and strategy in most organizations change every five years and yet BU leaders sponsor innovation projects. The mismatch is apparent, yet surprisingly prevalent in many organizations.
This, often, leads to one of two things:
- The venture is ‘managed’ by the BU sponsor to fit the needs of the current business, and minimize risk and disruption- choosing to fit into the BU strategy rather than looking for product-market fit.
- The BU sponsor gets promoted or moves elsewhere in the organization and her successor fails to see value in the venture and kills it.
This probably will not be a problem in cases where once-in-a-decade, paradigm-shifting technology, like AI, poses an existential risk to your company. In all other cases, the most obvious workaround for this pitfall is to change ‘culture’. But we know that takes a lot longer than one would like. Here are other some steps that you can take:
- Put ‘Market fit’ before ‘all other’ fits: If the objective of your ICV is breakthrough growth, especially using a specific technology, then deprioritize strategic fit. This may sound like a controversial statement but this advice has been followed successfully by at least two of the organizations in our report: the large chemicals player that allows the ICV to spin out, if it doesn’t find a willing strategic ‘buyer’ in-house; and the large materials manufacturer that has a separate fund allocated for ideas that are rejected by the core business.
- Invest ‘Patient Money’: It took BigTech decades to get to this point in AI – and billions in funding. Depending on the industry, the types of products, and research intensity, you need to set realistic expectations with your leadership team regarding time-to-profitability. If there’s a mismatch in expectations, you might be better off choosing an alternative format of innovation than ICVs.
This approach is exemplified by 3M leadership which tells its intrapreneurs:
“Our experience tells us that prior to its entry into the market, we do not really know how to anticipate the sales growth of a new product. We will [...] not say at the outset, ‘The idea must be capable of generating $50 million or $100 million per year in sales.’ Of course, we prefer larger businesses, but we will accept smaller businesses as entries into new fields.”
Why bother with ICVs at all?
Even if these pitfalls haven’t turned you off of internal venture building completely, you may still be wondering: since Microsoft seems to be coming out on top in the generative AI battle, doesn’t it mean that external innovation (through partnerships or equity investments etc) is the superior approach?
We think ICVs may well be the best approach in many cases and for many companies. As for the specific example we’ve considered here, it’s a little early to call a winner in the Google-Microsoft battle just yet- at the time of writing Google launched BARD to the US & UK markets. More generally, there are three main reasons why ICVs can still be worth doing and why many organizations continue to iterate and optimize this approach:
- Our focus, in this article, has been on the pitfalls of internal corporate venture building. But external innovation – whether through partnerships, equity investments or acquisitions – also has many well-documented pitfalls.
- With new technologies taking less time than ever before to gain a formidable market following (ChatGPT’s five days to 1 million users, as a case in point), the acquiring-external-innovation strategy is only going to get more expensive and your organization may not have the $13+ billion needed to build a partnership with a startup.
- Seeing their innovations having an impact on the market is a critical need for employees. The exodus of many Google AI engineers (59 of whom are at OpenAI) is proof of this. ICV offers a way to keep your best performing and ambitious talent engaged by offering growth opportunities that would not be as readily available through alternative innovation strategies.