What should an AI business strategy be, and how should it implement AI tooling? Many entities struggle, finding cost, security, and benefits failing to square. Problems like the inability to collect and mobilize internal data, and cost/workflow hurdles with years of work hold companies back. If desiring a custom implementation leveraging the power of AI and minimizing those pitfalls, Perficient can help.
However, the landscape shifted in two significant ways over the last couple months, with OpenAI releasing ChatGPT Enterprise, and Microsoft enacting strong legal protections for its tooling. This enables a practical, powerful strategy:
Turn it all on.
No Implementation, All Benefit
Data strategists argue that a company’s propriety data is important to integrate with AI. However, there are massive benefits from AI, regardless of integration.
ChatGPT had 100 million users in January, 2-3 months after release. Released soon after were Bard, Claude, and Bing. Those using AI — only the chatbots — could be closing in on 400 million.
The internet took eleven years to achieve 400 million users. This implies that AI is instantly useful, and the upfront work required to make it accessible has surpassed the inflection point. A Harvard study validates this hypothesis.
An examination from the university found that a group of consultants completed tasks 25% faster with 40% higher quality when using ChatGPT-4. With the flip of a switch, an employee gains a 25% productivity boost, and a near 50% increase in output quality. Importantly, ChatGPT did not have access to propriety data other than what the consultant gave it while prompting, meaning that this increase in productivity is not based on expensive data integration processes.
Security, Legality, Pricing
If the implementation and processing is not on premise, how secure is it, what kind of legal protections exist, and how much will it cost?
Security
Security is extremely important, and most major players in Enterprise AI capabilities promise the highest compliance and protocols available. Microsoft’s Azure cloud, and offerings like M365 are already industry standards and widely adopted. Copilot, Bing Chat Enterprise, and everything else AI-driven Microsoft offers is held to the same standard and kept within the Microsoft Compliance Boundary. ChatGPT for enterprise is certified to the same standard, offering SOC 2 compliance, AES 256 encryption and TLS 1.2+.
Neither Microsoft nor OpenAI keep the data or use the data to train. ‘Data’ encompasses both inputs and outputs.
Legality
Microsoft announced that they now legally protect everything their AI tools output. This includes Teams Copilot, Github Copilot, Bing Chat, and everything else Microsoft offers with AI. If a client or customer takes issue with something the AI created, organizations are not responsible, Microsoft is. OpenAI has not made promises around legal protection at the time of this writing.
Pricing
Currently, enabling the Microsoft stack comes at a price of 30/person/month. ChatGPT for Enterprise is unlikely to be much cheaper. Is this price quantifiably justified?
The Harvard study indicated that across a group of over 700 consultants, they averaged 25% higher output and 40% higher quality. As an average, this ensures that it does not take an expert user to achieve results, making the cost easily covered by the gained output.
Justification
The low end for the average billable consultant is $150. Assuming a lower utilization of 65%, this consultant would bill roughly $200,000 in a year. If they only managed to achieve a 15% boost from AI tooling, this translates to an additional $30,000 a year in productivity. This does not include the effect that an increase in quality would also have on billable rates and ability.
In contrast, the investment is $360 a year. Even if organization-wide access is not feasible, allowing individuals to opt in makes a valuable investment. The future of knowledge work is AI-assisted, and organizations that move first will find immediate advantage.
Closing Thoughts
Integrating new technology into an existing workforce is tough. AI enables companies to avoid that headache altogether, offering out of the box tools with enormous benefit.
Not all workers or tasks benefit at the same level from machine integration, but the data indicate rarely seen and immediate benefits for companies that can identify even a fraction of the workers and tasks that will. Unlike technologies in the past, the approach should not be “wait and see” but start and find.
If you enjoyed reading this, Drew Taylor often writes about AI, from using it in the workforce to the theory of singularity.
