Generative AI is a buzzword that is working its way into almost every conversation at the executive level and in a lot of ways for good reason. There’s a lot to be excited about with this technology, which over the period of a year gained traction with anywhere from 30% to 70% of the workforce using it depending on the industry and finding productivity gains in spades. But with all this discussion of Generative AI, is it a set of independent features you and your customers should expect in all of your products or is it a single cohesive system that you can curate, secure, and manage in a manner similar to traditional IT infrastructure? What should you expect if you embrace Generative AI from an implementation perspective, and how will that impact your business’s investments, staffing, and projections in this space? This blog post will try to address those questions and help identify some of the challenges as well as a solution in this space.
Some of the bigger questions IT departments face are, “Where does Generative AI fit into the existing infrastructure and software licensing ecosystem?” and, “Is Generative AI a feature or a system?” These questions are challenging as IT is expected to maintain compliance around security, logging, and the added pressure of appropriate usage and information guardrails.
Generative AI as a Feature
There are two major challenges with treating Generative AI as a feature, where each system or SaaS solution has its own Generative AI capabilities independent of each other: lack of holistic information and independent security solutions.
Holistic information can be the difference between Generative AI providing useful contributions with little effort by the employee vs. high-effort reviews and low-value add of Generative AI/Generative AI value realization. This is a similar situation to having two systems that are being integrated from two different vendors and being on support calls with each separately. Both sides know their own domain area, but getting the two to work together can be frustrating. This also applies to Generative AI as well.
From a security and risk perspective, Generative AI as a feature opens doors to challenges around security and having to lock down many different Generative AI workloads, and information ecosystems. Challenges like tracking the prompts used with the AI, training data set curation & cataloging, source information extraction, and guardrails around topic adherence / appropriate usage become major operational concerns. Each system must then either be left on its own with limited information, or integrated with your security ecosystem, and all SaaS information boundaries must be rigorously reviewed.
Treating Generative AI as a unique system allows operational teams distinct advantages, such as information security, training set management, reducing SaaS lock-in, and increasing the value realization of Generative AI. Amazon Q Business is a great example of Generative AI as a System, allowing all your various existing solutions to contribute to the Generative AI information pool and extend their existing functionality with Generative AI. Connecting your task-tracking solutions, process documentation, CRM, call center logs, and many other data sources can allow Amazon Q Business to leverage a variety of contextual information when answering questions.
Focusing on security first though, Amazon Q can do all of the above while still allowing IT operations to maintain tight control over which networks have access to it, what information the user has access to, guardrails around what types of questions can’t be asked, the ability to log all questions being asked (allowing for forensic analysis and AI/ML security monitoring), and various other operational advantages. As new systems are purchased, contracts can be more effectively reviewed for information usage criteria, and unlike other Generative AI chatbots, your data won’t be used to train the foundational model for Amazon Q.
Secondly, vendors offering Generative AI capabilities are partially working on building up customer habits around their Generative AI capabilities, which have a significant potential to both increase in cost over time, as well as be cut depending on the company’s ROI on them. Both of these mean that not only are you negotiating for the SaaS service, but also for the Generative AI feature costs, such as large software companies adding a new Generative AI feature which has a per-user cost just for writing emails. Centralizing this with Amazon Q Business can also stabilize your Generative AI costs to a single point of billing, allowing for more predictable costs over time. This also allows you to change vendors, such as ticketing systems, without losing efficiency with the Generative AI tooling your employees have access to.
Finally, we all expect Generative AI to have significant impacts on our ability to be more productive in a variety of tasks throughout the workforce, from reducing training to increasing efficiency. The problem with these expectations is that they require a lot of information about your business, how it operates, and past experiences in your business. The approach of Generative AI as a System allows all of your data to be pulled together, including business process data, and answer questions, frame statements, or break down work based on past experiences. Systems such as Amazon Q Business also allow for integrations back into those original systems. This enables your team to not only leverage Amazon Q Business to answer questions, break down work, etc. but also to directly have Amazon Q Business then push that information back into your systems using either out-of-the-box integrations or custom integrations.
Amazon Q is a generative AI-powered assistant for leveraging companies’ data and accelerating software development. There are five Amazon Q products, including Amazon Q Business, Amazon Q Developer, Amazon Q in QuickSight, Amazon Q in Connect, and Amazon Q in AWS Supply Chain.
The first product, Amazon Q Business, is a generative AI–powered assistant that can answer users’ questions, provide summaries, generate content, and securely complete tasks based on data and information in an organization’s systems. It empowers employees to be more creative, data-driven, efficient, prepared, and productive. It’s secure and private by design – safety, and security was intentionally built into the product and is a top priority. Amazon Q understands and respects existing governance identities, roles, and permissions, personalizing interactions accordingly. If a user doesn't have permission to access certain data without Amazon Q, they cannot access it using Amazon Q either. It easily and securely connects to over 40 commonly used business tools, such as Salesforce, ServiceNow, Gmail, Microsoft Exchange, intranets, Amazon Simple Storage Service (Amazon S3), and many others. Simply point Amazon Q at your data and it will search, summarize logically, analyze trends, and engage in dialogue with end users about the data. This helps business users make sense of their data no matter where it resides in their organization.
It's incredibly easy to get started and operationalize Amazon Q Business, providing Generative AI as a system to derive immediate value.
Corporate buy-in can be challenging with new technology, especially if there’s a large upfront time commitment to realizing the value of that technology. The longer implementations take, the more there is the questioning of whether the investment is worth it, the potential to lose key stakeholders, and even the potential that the capitalization window passes you by or is replaced by a packaged product. While waiting for a solution could mean that the opportunity to gain market share decreases due to a lack of action, or a delay in the realization of millions of dollars in missed revenue potential, reduced organizational complexity, increased revenue-producing productivity, and higher market share realization.
From Trek10’s experience with Generative AI, organizational buy-in comes from the crawl, walk, run approach. In the crawl stage, you want extremely fast speed to value, handling more generic use cases that improve the efficiency of your workforce at a low implementation cost. At the walk stage, you and your team have progressed to understanding where Generative AI can impact business outcomes, and you invest in cleaning your data and find value in analytics toward the end goal use-case. Finally, you create your use-case-specific Generative AI solution reducing costs, while increasing value realization.
Amazon Q Business helps find the immediate value in the crawl stage, by providing a packaged Generative AI service, which has a low implementation cost, high security/ops friendly, and produces value quickly (2 - 6 weeks of implementation for a lot of use-cases). This creates a low-risk, high-reward scenario where things such as user interface, infrastructure, and network security are simple steps in the implementation, while control over your use case and information is high and integrates with your existing identity provider.
Amazon Q Business also decreases the amount of specialized workers needed in your organization if configured correctly. Generative AI without Amazon Q can require a significant investment in AI/ML Ops, Data Science, Security teams that are Generative AI aware, operational playbooks, and more. With Amazon Q Business the focus turns to Data Ops and an introductory level Generative AI aware security team. This reduction in specialization helps with early-stage success with early Generative AI use cases putting less stress on organizational build-out.
By using Amazon Q Business as a launching pad, you end up with having demonstrable Generative AI capabilities to your executive team and board, while also enabling your internal teams to start realizing the value of Generative AI responsibly.
Amazon Q Business has a lot of upsides when it comes to ease of implementation, but there are also quite a few considerations when it comes to making sure you have it set up correctly, in alignment with best practices, knowing which connectors are mature vs. which aren’t, and how to get past specific issues with some of the initial implementations and pre-prompting in order to get the best experience possible out of Amazon Q Business. This is where Trek10 in collaboration with Amazon is offering up to $25K in free professional service work to implement Amazon Q Business through the 2024 calendar year. This amount is typically enough to get your business set up with Amazon Q Business in the right way and ready for production.
The $25K in funding can apply to almost any use-case including a few marketplace use-cases which we’ve found as good starting points.
The Trek10 ATTAIN: Employee HR & Policy Chatbot with Amazon Q Business offering focuses on the internal use-case of supporting employees and management with HR policies, employee review, connecting training course information, and more. By connecting these different sources of information, your employees can easily ask HR-related questions and get answers quickly without having to take time away from your HR department. The following link is an example of how Amazon Q Business can help the Annual Review Cycle.
The Trek10 ATTAIN Retail: Internal Support Chatbot, Trek10 ATTAIN Financial Services: Internal Support Chatbot Accelerator, and Trek10 ATTAIN Travel & Hospitality: Internal Support Chatbot focus on your support team’s ability to be more effective in helping your organization and customers by augmenting them with Generative AI. This approach can often reduce training costs, and increase correct remediation steps by leveraging Amazon Q Business to pull together both legacy and emerging information in support of your organization and customers.
There are a variety of other use cases for Amazon Q Business, and Trek10 can help your team not only get started but can help train them on how to manage Amazon Q Business implementations. Trek10 can also help offload the management of these systems from your team either initially or ongoing with our Trek10 Cloud Operating Platform.
Even if you’re in the early stages of evaluating Generative AI for your business, Trek10 is here to help and would be happy to hop on a call with you and talk through your use case. Reach out to info@trek10.com at any time to engage with our team.
This article was written in collaboration by Brenden Judson (Trek10), Nikody Keating (Trek10), and William Lorenz (AWS).
Learn how Claude3 Opus, now available on Amazon Bedrock, outperforms its peers on common evaluation benchmarks for AI systems.
