Speed 2: Control Your Cruise

Good news regarding my first article in this series, McKinsey agrees as of a LiveWebinar this morning that will shortly be replayable.

One of the most important assertions in my article was that Gen AI makes teams so productive that it puts a burden on leaders to set direction and update priorities at a pace they are likely unprepared for.?

The McKinsey analysis shows this rapid productivity acceleration is especially clear for Software Engineering and for Sales & Marketing, each representing $1.2 trillion of the $4.4 trillion in performance boost they can identify from Gen AI.

The challenge for leading adopters of Gen AI in software development is feeding priorities to their super-productive developers fast enough. At Azure Build, everyone including the CEO made a little joke about how Microsoft woke up on January 1, 2023 and decided to start launching co-pilot products at a furious pace.? They did because, suddenly, they could.

There was also a warning in my article that AI-leading companies who have adapted and accelerated to this pace have a greater capacity to execute initiatives that might include coming after your customers.?

Another way to say that is from the McKinsey partner, Dr. Michael Chui, who pointed out today, that acceleration opportunities create the potential for greater competitive distance between the leaders and the laggards.

A speed boat to where?

Expanding on the pressure on leaders, here is the pattern that starts in software development and will radiate outward to all knowledge work.

1. What over how. Gen AI makes knowing what you are trying to do more important than ever because knowing how to do virtually anything is now vastly easier than ever.
2. Next is now. Gen AI radically shortens the cycle time between what you are trying to do now and what you want to do next because your teams now have or are about to have an incredibly potent co-pilot for nearly every task.
3. Prioritization pace. Outperforming expectations increasingly will mean successfully setting priorities at the accelerated pace your teams can execute.
4. Instrumented climb. Superior prioritization flows from superior ideation and collaboration combined with superior, data-rich, after-action analysis.

The alternative to acceleration is always available, ie to take the productivity to the bottom line. Many will choose to reduce headcount and live with the risk being passed by those who execute accelerated prioritization.

Marketing at GPT speed

As previewed, let me give you an example from the marketing world.?

On May 23, Adobe announced Firefly which is Gen AI for nearly-instantly creating multiple generated image options from text prompts.?

We’ve seen this with Midjourney and Dall-E, but Adobe has “Fill” features to expand images and delete or add items with correct reflections, shadows, lighting, etc. So you can use this for new or existing photos. images or PDFs.? And Adobe is vouching for its rights to have used all the images to train the model such that there will be no downstream copyright infringement claims.

From May 23 to June 9, users created over 200 million unique assets. Maybe that was not marketing, but just for fun.

Less than two weeks later on June 9, Adobe released a new version of Adobe Express which is their all-in-one content generation application that marketers use to generate compelling visual content for all kinds of sales and marketing campaigns.

Firefly is now embedded into all content creation workflows - videos, designs, and documents formatted for all kinds of web and social media destinations. A few lines of input and a few clicks.

I’m gonna take the rest of the day

The demo giver, Paul, at the Adobe Summit in EMEA on June 9 even ends with the bottom line, “I’m gonna take the rest of the day off.”??

Is he? Or is he going to start tomorrow’s work? Or is he going to tinker all day with what he created to make it 1% better?

On the back of that thoroughly convincing proof point, Adobe announced they will bring in Gen AI from multiple large-language models under one brand name, Sensei AI, across their applications starting now in beta in some cases. Marketers committed to Adobe end-to-end can expect to:

• Generate responses to customer questions for sales, and then summarize the interaction.?
• Create and modify marketing copy in the right tone, which they can even train on their existing content for brand consistency.?
• Use an AI assistant to let them pose natural language questions to access Adobe’s wealth of best practices for using Adobe technology.?
• Make natural language queries for customer analytics.?
• Automatically generate key takeaways and add meaningful captions to charts and graphs produced by Adobe applications.

On June 13, Salesforce previewed a similar strategy. This will continue every week from other vendors.

Predictive vs Generative

Predictive AI is specialized, for expert use. GenAI is horizontal, pervasive, and accessible.

Whether you watch developers at Microsoft using Github co-pilots or Paul from Adobe Express building beautiful content, you come to the same three conclusions.??

1. That was fast.?
2. That was accessible, ie easier to learn and really easy to do.??
3. You better know what you want (else you will get a lot of stuff all over the yard.)

But didn’t SaaS eat BPR?

I keep prescribing business process reengineering (BPR) to capitalize on this speed.

A critical question, though, is whether some time ago we essentially replaced BPR with adopting the right software for the job. ? Is it true that Digitization = SaaS Adoption = Best Business Practices?

After all, the era of PC | Client Server | ERP ended with most of the focus on implementing ERP and Supply Chain software.??

The era of Web | E-commerce | Big Data ended with most of the focus on implementing E-commerce that with a little AI could personalize what a buyer sees, along with ERP’s children: CRM and HR systems.??

And the age of Elastic Cloud | Mobile | and X as a Service kicked off with Sales and Marketing SaaS and then saw an explosion of Enterprise SaaS, Cloud-Services catalog offerings, and Mobile Apps.

Banquet, feast, or avalanche?

By explosion, I mean that companies now use hundreds of SaaS applications.?

Bettercloud’s 2023 survey says the average number of SaaS applications an enterprise uses has gone from 8 in 2015 to 130 in 2022. A CAGR of nearly 50%.? Within that, you find 190 applications for companies with 750 to 2,999 employees - a size that means there is really just one business inside that company, which is using 190 SaaS applications.

Surveying enterprises of 1,000+ employees only, Mulesoft and Deloitte Digital say in 2023 the average enterprise uses 1,061 SaaS applications up over 25% in two years from 843 in 2021.

In addition, enterprises use several dozen cloud services from AWS, Azure, Google and others. Add in a handful of mobile enterprise apps for email, collaboration, and projects.

This year, each existing SaaS, cloud service, and mobile app will get many Gen AI-based enhancements.

The answer is: Avalanche

Gen AI introduces a whole new category of SaaS, Cloud Services, and Mobile apps. In fact, some are wholly new, and some are enhancements of existing ones.

Take a moment to consider this graphic. MAD FirstMark ML/AI? I didn’t include the graphic because you would not be able to make out the companies and probably not even the categories because there are so many.

Digital without Transformation

The vast majority of companies equate the adoption of SaaS with digital transformation. It is certainly digitization. The question now is whether it is transformation.

Borrowing from an idea from a Microsoft Fellow called Steven Batiche, one way to sort how GenAI affects your business is into three broad groups as follows.

Alongside: Co-Pilots. Applications that include a ChatGPT-like sidebar, are further enhanced with optional plug-ins.? Co-pilots make sense of what you’re seeing and doing with the application, suggest actions, and even do tasks for you. Microsoft Azure is off to the races here especially for developers. Other systems used by experts, engineers, and scientists are also getting these. Salesforce for example is using Einstein GPT as a co-pilot.

Inside: AI-Applications.? Applications that use AI to enable users to create results that only super-power-users could do before. ChatGPT, Bing w GPT, and Bard are examples. Google has just shown this to be their primary AI focus at their conference, ie to make search, maps, docs, photo editing, their smartphone, etc amazingly better experiences auto-magically with AI inside. Check out the fly-through preview for maps.? Adobe is doing this. You can expect Apple to continue to focus here when they finally unleash.

Outside: Orchestrating across applications. Pulling data from multiple applications, workflows, and other internal and external data stores related to a business process gives GenAI the data to create a coherent view or even a simulation so it can help you analyze and even suggest improved actions in any combination of those workflows.? Orchestrating.?

Salesforce, for example, announced that building on predictive AI they trained while respecting their customers' data, they are enhancing each element and across a whole Salesforce-powered Revenue generation process with Gen AI, responsibly gathering and anonymizing all the data necessary to train, robustly inform, and effectively prompt GPT models for each users job.

It is this last one, orchestration, that will determine transformation and the winners.

Pile or Stack?

Looking at those who are far down the road like software developers, you find that effective leaders don’t have a pile of software tools across their teams. Rather they have a carefully selected and ever-improving stack that the whole team has agreed to use.

At some layers of the stack, they agree to let each person or team make a choice. But overall, it is coherent, designed for output, instrumented to measure component and overall results, and yet forever undergoing improvements. And critically, the leaders make data flow across the tools supplying each step and measuring the entirety of the business process.

Whenever two development leaders meet and start talking shop, they go through a ritual of explaining some key updates to their stack and asking each other why they chose A or B and what they were doing about prior choices X or Y.

For new acquaintances, this is a very quick test as to whether the other person actually is leading development. If they can’t engage in this ritual, they aren’t a development leader.?

One part digital, one part transformation

Across alongside, inside and outside categories, the same two-step pattern is clear. Step one is choosing a great technology stack and making it a coherent whole with data flowing across it. Step two is engineering how you evaluate and set priorities so that you get accelerated and higher-quality results.

Alongside. Your challenge with co-pilots is prioritizing. In general, each employee will get the benefit of AI and continuous up-skilling based on choosing the right software to adopt. Many of these tools will even pull in data from other sources.

However, building a well-informed, continuous prioritization backlog to keep your teams pointed in the direction as they accelerate is critical.

Inside. Your challenge with AI applications is measurable quality. Ironically, if your experts get their tasks done a lot faster and at an expert level, they may just choose to do more iterations of the same tasks. Guido Jouret calls this: Supply creates its own demand.?

So, high on the list for reengineering is instrumented sprints in functions outside of software development. The idea is to measure the output or impact of coordinated rounds of work. Curate, through team collaboration, improvement actions to include in the next round. Don’t change the underlying products except via the next sprint, patches excepted.

Outside. Your challenge with cross-application orchestration is that you have to do it. Orchestration is pretty much your job as the business leader

In some functions like Marketing and Sales especially, you may be able to get some of your reengineering done by adopting a major integrated AI-enhanced platform like Adobe SenseiAI and Salesforce GPT.? You get data integration across tools, hopefully, and some defensive capabilities done for you too.

Still, your focus is executing two things in parallel. One, getting really good at using the platforms, and two, re-engineering your priority-setting cadence to achieve the instrumented sprints mentioned above.??

Build your own adventure

Those unified, all-in-one platform decisions are impractical for many who have gone down other paths or are operating on shoestring budgets.? For many functions and industries, no such off-the-shelf end-to-end automation exists.?

If McKinsey is right, there is a lot of value available outside of the software industry. $400 to $600 billion in Retail & CPG, $200 to $340 billion in Banking, and $60 to $110 billion in Life Sciences, for example. So too for functions outside Sales and Marketing.

One great attribute of Gen AI is that it can translate any structure to any other. So there is no technical barrier to making data flow across tasks, applications, departments, and other silos. But you have to ensure that happens. Restrictions apply, e.g. privacy.

For functions that are truly unique to your company, you have to take the blue pill: assemble your ever-improving stack from a vast sea of choices with data flowing, and reengineer your priority-setting processes for those functions.

Next up

In subsequent articles, I will go deeper into use cases for instrumenting sprints in Marketing functions to illustrate what is possible and essential in multiple business processes. I will expand on the lessons learned for making data flow across digital silos.

And I will explore some of the advanced opportunities like Digital Twins for your business process, which today is the domain of the most sophisticated engineering groups.? Tomorrow is a different story and coming at us fast.

Please let me know what you think.