Artificial Apotheosis

The Year 2023 was typified by consumers and companies alike becoming enthralled with the possibilities of artificial intelligence: from generative AI that unlocked new levels of creativity and productivity, to delegative and autonomous workflows that are ushering in scale like never before seen. Is 2024 the year business becomes serious about synthetics

In October 2022, signs of the Singularity jolted a technology sector that had been lulled into a post-pandemic malaise with excitement and energy. And yet throughout 2023, both the promise of artificial intelligence and the disruption of generative content resulted in the mass-disintermediation of organizations from their customers and constituents. Every software provider desperately rolled out an AI chatbot or API service in what could only be described as a disorganized and chaotic effort at not missing out on the marketing potential of the craze. None of these were special - unique or differentiated - and unsurprisingly, all have failed.

In the wake of mass layoffs stemming from the software industry's failures to truly reinvent themselves through artificial intelligence, and the rising risks borne of this radical moment as technology companies underwent disingenuous rebranding of their older automation tools as "AI," analysts are only now beginning to lean into the seriousness of what is at stake. The Catastrophe of OpenAI's near-collapse just weeks ago crystalized how so many sectors had begun to architect the facade of a future built upon ChatGPT, what IBM's Senior Vice President of Research Darío Gil warned earlier in the summer was "reducing the enterprise AI strategy to an API call." The real transformative work of reinventing businesses will require far more industry and courage in the face of terrifying and thrilling odds.

Confronting the Challenge

Software-as-a-Service, or SaaS, which primarily dominates the digital age following the dot-com collapse at the start of the millennium, is confronted by its own demise in the face of artificial intelligence. SaaS products - from website builders and content management systems (CMS) to customer relationship management (CRM) systems and every application powering virtually every facet of our radically connected world - are little more than front-end interfaces that simplify the complexity that business users and retail consumers face in performing computational tasks. Whether it's shopping online, managing your bank account, or posting updates for your friends and family, people have reluctantly grown to tolerate (but not truly love) their experience with websites. In fact, an entire "digital detox" industry has arisen led by proponents such as Ariana Huffington, urging people to disconnect from the dangers of social media and the excesses of ecommerce. Nevertheless, SaaS has prevailed because other efforts to make the user experience more compelling - more timely, valuable, and relevant - have failed to catch on.

Artificial Intelligence (AI), especially generative AI (genAI) and the fast-arriving artificial general intelligence (AGI), are more than just the latest online fad. While the successive novelties of cryptocurrency, blockchain, and non-fungible tokens (NFTs) have each seen their red-hot ascent and terminal "winters," artificial intelligence is presenting a new and unprecedented interface to the human experience. It is for this reason that innovation analysts describe it as the most important invention since the printing press, and something that Microsoft's own engineers opine spells the end of software-as-a-service. If people are able to live, work, and play without having to avail themselves of outdated web-based interfaces, and can instead engage in human-like conversations and breathtakingly realistic creations that feel intuitive and natural, it spells certain doom for a SaaS industry that awakens to its own obsolescence. It's little wonder why the CEO of largest SaaS company Salesforce so cravenly tried to recruit OpenAI's engineers in what Fortune magazine called a "feeding frenzy." The path to prosperity in the software industry runs through the heart of artificial intelligence.

Architecting the Apotheosis

Generative AI has launched a massive wrecking ball into the legal and economic framework underpinning intellectual property, with most regulatory bodies struggling to even comprehend (let alone, come to terms with) the very nature of synthetics. In the most recent example of artists, writers, and musicians filing legal challenges to artificial intelligence, the New York Times sued OpenAI just days ago for using its content to train the models that are now used by hundreds of millions of people around the world. Furthermore, the very models built upon decades of data freely available on the Internet face continual collapse as OpenAI's GPT and DALL-E, Midjourney, Runway, Pika, and many of the other nascent startups are striving to scale by training off of user-generated synthetic data to the point of hallucinogenic dysfunction, rendering their future suitability questionable.

The solution is straight-forward, argues Gil: enterprises must architect their artificial intelligence around their proprietary and private data. Rather than outsourcing generative AI and business workflows to unaccountable third-parties, or worse yet willingly training the models of companies' would-be competitors, organizations will be better served through a five-step architecture process to building an AI-driven future for their business products and services as SaaS declines in the months and years to come. 

  1. Prepare Data
    The most valuable commodity that businesses and individuals have is their data, the underlying information that drives proprietary and private knowledge. Data must be prepared, across the public cloud, private systems, and on premises data lakes and warehouses, and made readable by machines. This preparation process includes a rigorous clean-up and categorization process to ensure the quality of the data is suitable for use.

  2. Train
    Once properly filtered, organized, and tagged, the data is then suitable for use in training a type of model, a behavioral framework used by artificial intelligence, to understand the proprietary data's value, significance, and ultimately, human-like meaning. Whether non-generative and encoded, encoded-decoded, decoded-only, novel-identified, or densely associative and expressive architecture, each model framework offers specific benefits to their corresponding enterprise workflows.

  3. Validate
    After the intense training process, that tokenizes proprietary data into trillions of relationships, the model is ready for benchmarking to validate the model's quality across a wide range of metrics. These metrics are specific to the intended business application: from customer support, to data analysis, and even executive autonomous functioning and decision-making. Only models that pass the thresholds are suitable for use.

  4. Tune
    Adapting the model to downstream paths is the basis for the productivity gains inherent in artificial intelligence. Ensuring that the model can replace and far outperform other experiences (such as a SaaS web-based interface) requires leveraging the model against a specific set of expected results, which will in turn produce a set of ideal prompts to use for the model and its data. 

  5. Deploy
    Specialized models that have been fine-tuned and validated for experiences are ready as solutions to be deployed for use by end-users. The more specialized and highly-tuned the model against privileged and proprietary data sets, the more impressive and impactful generative and autonomous data systems can scale for applications. Continuous monitoring of both data and the models built upon them remains essence for proper long-term governance to ensure its performance.

Artificial Intelligence is entering its Apotheosis - the moment where divine qualities are ascribed to its potential, but its own anticipated fall from grace can occur if humans mismanage this incredible power. Enterprises have critical choices, much as businesses did at the start of the industrial revolution in the eighteenth century, and a misstep (or indecisiveness) can lead to irrelevance and insolvency. We can help. Embarking on a transformation around artificial intelligence requires trusted partnership grounded in experience and results to which our clients can attest.