Large language models (LLMs) are everywhere. They do everything. They scare everyone – or at least some of us. Now what? They will become Generative-as-a-Service (GaaS) cloud “products” in exactly the same way all “as-a-service” products and services are offered. The major cloud providers – “Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform (GCP), Alibaba Cloud, Oracle Cloud, IBM Cloud (Kyndryl), Tencent Cloud, OVHcloud, DigitalOcean, and Linode (owned by Akamai)” – will all develop, partner or acquire their generative AI capabilities and offer them as services. There will also be ecosystems around all of these tools exactly the same way ecosystems exist around all of the major enterprise infrastructure and applications that power every company on the planet. Google is in the generative AI (GAI) arms race. AWS is too. IBM is of course in the race. Microsoft has the lead.
So let’s look at LLMs like they were ERP, CRM or DBMS (does anyone actually still use that acronym?) tools, and how companies make decisions about what tool to use, how to use them and how to apply them to real problems.
Are We There Yet?
No, we’re not. Will we get there? Absolutely. Timeframe? 2-3 years. The productization of LLMs/generative AI (GAI) is well underway. Access to premium/business accounts is step one. Once the dust settles on this first wave of LLMs (2022-2023), we’ll see an arms race predicated on both capabilities and cost-effectiveness. ROI-, OKR-, KPI- and CMM-documented use cases will help companies decide what to do. The use cases will spread across key functions and vertical industries. Companies anxious to understand how they can exploit GAI will turn to these metrics and the use cases to conduct internal due diligence around adoption. Once that step is completed, and there appears to be promise, next steps will be taken.
Preparing for the Inevitable
What will you do with LLMs when they’ve been fully productized and then even commoditized? The next step is to identify and describe the business functions and processes that might benefit from GAI. This is not an easy process mostly because most companies do not have process inventories or have “mined” the processes to identify the features most amenable to GAI. It’s also complicated because whole business models (comprised of many processes) may be GAI targets. Take marketing, for example:
“Professionals who work in a corporation’s marketing and promotion departments seek to get the attention of key potential audiences through advertising. Promotions are targeted to certain audiences and may involve celebrity endorsements, catchy phrases or slogans, memorable packaging or graphic designs, and overall media exposure.”
Marketing processes, services and products?
- “Marketing refers to all activities a company does to promote and sell products or services to consumers.
- Marketing makes use of the ‘marketing mix,’ also known as the four Ps – product, price, place, and promotion.
- Marketing used to be centered around traditional marketing techniques including television, radio, mail, and word-of-mouth strategies.
- Though traditional marketing is still prevalent, digital marketing now allows companies to engage in newsletter, social media, affiliate, and content marketing strategies.
- At its core, marketing seeks to take a product or service, identify its ideal customers, and draw the customers’ attention to the product or service available.”
Which of these processes can benefit from GAI? Can GAI develop marketing campaigns? Can it write press releases? Can it target customers? Yes, it can. But with what quality? Can it develop campaigns, press releases and strategic customer targeting as well as marketing professionals? This is the ongoing assessment that will define adoption. It will also determine how the hybrid/partnership role that LLMs will evolve until they assume leadership roles which they will in specific – especially well-bounded – domains.
HR? What are its processes?
- “Human resources (HR) is the division of a business responsible for finding, recruiting, screening, and training job applicants.
- HR departments also handle employee compensation, benefits, and terminations.
- Human resource management (HRM) strategies focus on actively advancing and improving an organization’s workforce with the long-term goal of improving the organization itself.
- HR departments must keep up to date with laws that can affect the company and its employees.
- Many companies have moved traditional HR administrative duties such as payroll and benefits to outside vendors.”
How will GAI improve, replace, automate or reimagine HR processes? In time? Pretty easily. (And let’s not forget that good ol fashioned machine learning can improve, replace, automate or reimagine many processes.)
This drill will play out across the globe. The process/technology – in this case, GAI – “matching” process will preoccupy CIOs, CTOs, CEOs, CMOs, CFOs – and all of the chiefs pretty much forever, as it will regarding the process/technology matching with other emerging technologies.
What else? Companies need to track GAI closely – including what their competitors are doing with the technology. Since the technology will grow beyond Moore’s Law, companies might consider creating Task Forces and even Centers of Excellence to track GAI and its application potential. Chief AI Officers? Maybe, if there’s room for yet another Chief.
Clouds & Ecosystems
All of the cloud providers will enable the above steps. There will also be a GAI ecosystem, which will consist of a:
“ … network of interconnected digital technologies, platforms, and services that interact with each other to create value for businesses and consumers.”
Ecosystems also include an array of vendors constantly improving their products and services, as well as new entrants seeking to disrupt the incumbents.
Consultants will offer advice around multi-LLM management, LLM security, prompt engineering and how to avoid (or at least identify) LMM “hallucination.” No doubt the consultancies will create other GAI services they can sell.
Like all new technology ecosystems, this one will suffer from talent shortages. Universities will scramble to develop new courses and degrees in GAI, but will lag what the consultancies will offer. (Curriculum committees cannot move fast enough to keep up with the field.)
Business as Usual?
Despite how many pauses occur, the development of GAI will proceed. It will begin as new technology, expand in the cloud (with the usual suspects) and take its seat among the major technologies that have contributed to all things digital. It will look and feel like ERP, CRM and DBMS applications complete with its own ecosystem. The number of venture investments will eventually slow as the number of use cases expands.
But GAI is different. While it’s tempting to go down the “what about (cars, planes submarines and Uber)?” path that were presented as disruptive technological leaps, GAI is different simply because of the breadth of its application potential and its ability to grow without scheduled maintenance. It’s also different because of the application opportunities (or threats, depending on where you sit) which are primarily knowledge-based:
“The U.S. number of fully and remote and hybrid knowledge workers will account for 71% of the U.S. workforce in 2023. In the U.K., fully remote and hybrid knowledge workers will represent 67% of its workforce in 2023.”
So pay attention to LLaMa, ChatGPT, Bard, Co-Pilot and all the others, and how large language models will become huge cloud services with massive ecosystems – and start identifying the processes that GAI can help. Chances are good that GAI can do much more than just assist.
Read the full article here
