Key Considerations in Selection of a GenAI Solution

Many CEOs and boards are asking questions about how their organization will use GenAI to meet and exceed their business objectives. Many enterprises plan to align with a platform or product to meet these expectations. In a fast-moving GenAI landscape, what are the critical criteria for selecting a platform or product to work with? This article attempts to provide guidelines for the same.

GenAI is a fast-evolving field with new LLMs, models, frameworks, use cases, and new products/platform developments coming out almost every week. Given the novelty of making product/platform decisions for GenAI– some guidelines might help.

Here are the top nine considerations for choosing the right GenAI solution for your organization.

Does the product/platform align with your use case?

GenAI technology can do various tasks. Here are some use cases that seem to be getting prominence within enterprises – both in terms of interest and availability of solutions. Please note that there might be some overlap between these use cases, but there is enough distinction to treat these differently.

• Enterprise Knowledge management: Content discovery from enterprise knowledge sources  
• Individual productivity tools: Help an employee by providing content summarization, creation, and search/answer based on existing content (your Outlook, Presentations, Documents, Excels, etc.)  
• Code creation solutions: These solutions are Meant for developers; they can write new code quickly based on existing code.  
• Customer Service: Chat-based solutions for customers typically integrated with CRM  
• Employee-facing Service Desk: Help employees with quick resolution of any issue or help  
• Marketing support solutions: Marketing communication optimization, content creation etc.  
• Industry-specific use cases: Pharmacy, medical data, supply chain optimization, etc.    

Key questions to ask

Question 1: Can a single platform suffice for all my use cases, or will I buy different solutions?  

Our view: Each of these use cases might look similar at a high level but have started to diverge regarding features/functionality needed. For example – let us know the knowledge management use case.  

Below are some features/functionalities needed for the knowledge management use cases.  

• Employees can get access to enterprise knowledge by asking the GenAI bot (typically integrated with conversational platforms like Microsoft Teams).  
• Ability to connect to varied knowledge sources (Websites, SharePoint. Confluence, documents, SNOW knowledge articles, etc.)  
• The ability for content restriction based on attributes/metadata  
• Ability to set prioritization sequence between knowledge sources  
• Ability to resolve older/non-current knowledge and other conflicts  
• Ability to give non-generated responses in certain situations (Scripted answers which are not generated)  
• Ability to provide videos or images as answers, etc.

As you can see, a lot of specialization is needed to be an excellent enterprise knowledge solution – and this is just the beginning of a GenAI-based enterprise knowledge management journey. As evolution continues – one could expect more specialization.

At the same time – a few use cases are adjacent enough for a single product to cover well. For example, Enterprise knowledge management and Employee support connect quite a few strings. For employee support, we also recommend checking out features like Live Chat or ‘Human-in-the-Loop' or HITL function – since intervention from support agents can be called on-demand by the employees.  

Question 2: can I buy a platform and build all the use cases using the platform?  

Our view: The aspiration to build all the use cases on a single platform might also result in some solutions that are individually sub-optimal. Building specialty IT platforms has always been very expensive and even more so to maintain.  

Question 3: Should I wait for existing products to solve for some of these use cases by incorporating GenAI in their existing product?

Our view: This can sometimes work, and you should ask your existing product vendors for their GenAI roadmap. However, please take the roadmaps with a pinch of salt. If the vendor is extremely slow in introducing GenAI features at the current time, then you might be dealing with a laggard in the market. You might have to look for alternative/emerging vendors that have a stronger GenAI approach.  

GenAI architecture & algorithms

The architecture of a GenAI solution is a critical differentiator. Based on the current wisdom/trends – a RAG-based architecture is considered a superior option. However, being a RAG-based architecture does not guarantee a great solution. For example, the algorithm with which RAG-based architecture creates 'chunks' for 'vectors' – will be different for different solutions – and can make a lot of difference in the outcomes of generated responses. This is what Rezolve.ai does.

Rezolve.ai employs a proprietary RAG-based algorithm that uses ‘chunking’ to power its vector-based data. This is then coupled with the GenAI engine to deliver a high-powered solution that banks on accuracy and reliability. You can read more about RAG and Chunking in our whitepaper.

To provide another analogy for architecture and algorithms, Google did so well in early 2000 because its search engine was much better optimized than its competitors. People quickly found out why they like Google over other search engines because of its precise results, speed, and user experience.

Recommendations

Check if your platform/product vendor can explain their current architecture. Also, play with the product to ascertain that the quality and speed of responses are per your expectations. You can check for your own solution against the pointers listed below.

Configurability: Most enterprises expect products that can be easily configured. GenAI applications should not be any different. However, an excellent question to ask is – how often will I have to tweak the configurations? What expertise does my organization need to perform these configurations and fine tunings? Do I need to hire prompt engineers to write prompts?

Our view: Rezolve.ai recommends easy configurability. Prompt engineering is a complex art and affects not only the quality of responses – but also the security of the solution (prevent hacking). We do not see many organizations venturing into writing their own prompts there. Product companies will be able to provide great prompts that work best for their product/use case.

Pilot Project: A test run for GenAI solutions, at the pilot scale, is highly recommended. Many companies are focused on implementing the solution right away while its long-term implications, cost-benefit ratio, and lock-in commitment are still uncertain. Before deploying the solution on full scale, a high-powered user pilot rollout is strongly recommended.

Our view: Rezolve.ai allows its customers to roll out a GenAI pilot project seamlessly. Not only is this helpful in demonstrating the proof-of-value of the solution for the organization’s use cases, but it is also a risk-free approach. Since Rezolve.ai offers a 2-week free trial, organizations can collect the user and stakeholder feedback on the utility of this modern GenAI ITSM system and proceed forward with confidence and certainty.  

Security & privacy

Evaluating a GenAI solution for enterprise security and data privacy is crucial to protect sensitive information. To make sure that a GenAI solution does not make your confidential knowledge the part of internet’s knowledge, ask these questions;

Question 1: Is my data safe? Will my data be used to train an LLM?

Our view: All reasonable solutions use enterprise APIs with guarantees about data security – especially the first two questions.

Question 2: Can my PII be stopped from reaching an LLM?

Our view: To address this issue, a DLP layer is needed. Please check if your GenAI solution has a built-in PII layer.

Question 3: Can your product use my enterprise OpenAI API instead of your own?

Our view: Some solutions offer using your enterprise's OpenAI API for further assurances.

Question 4: How can I be sure that our enterprise knowledge is being shared with the people for whom it is intended?

Our view: This is a crucial issue that needs to get the attention it deserves. Many platforms' solutions learn from entire SharePoint or similar significant knowledge sources – and apply the permissions on knowledge based on existing SharePoint permissions. It is assumed that existing permissions were set correctly – or the correct documents are in the proper folders. However – as we all know – that is not likely to be true in all cases. This can lead to unintended sharing of information. Therefore, Rezolve.ai does not recommend sharing the entire SharePoint – a subset of folders or newly created folders with specific information. Our approach and an enlightened information architecture will allow you to avoid any information-sharing risk.  

Recommendation: Please check with your product vendor on their approach to permissions, etc.

Actions & GenAI-as-a-middleware

GenAI solutions can do much more than provide answers. GenAI solutions can take action. This is very important as providing answers or summarization is only part of the bigger automation jigsaw puzzle. Actions can be as simple as creating a ticket/case and as complex as large workflows involving several organizational roles, applications, departments, or approvers. Typically, these actions need GenAI solutions to integrate with other enterprise applications.

These integrations are usually achieved through an external middleware or an internal integration hub. Given how central a role GenAI solutions are likely to take, the integration hub of GenAI middleware is likely to become a GenAI middleware glue that connects different processes and products.

The key discovery questions to ask here include:

• Does your GenAI solution have the ability to take action?  
• How mature is its orchestration layer? Does it have a studio?  
• Is the workflow or orchestration layer part of the product – or external application/module?
• Does it have auditable approvals built-in seamlessly?
• Are there some integrations that come out of the box?

Our view: Most use cases will need smart actions based on a high level of autonomy. GenAI solutions allow adding actions and automation without spending too much time and effort (famous low-code/no-code studios). Please request your product vendor for details.  

Key Considerations in Selection of a GenAI Solution-Ebook

Discover essential factors for selecting the right Generative AI (GenAI) solution tailored to your organization's needs in this eBook.

Get the free guide

The ROI of GenAI for your company

ROI is pivotal in investing in any technology, or the use cases it promises. ROI is typically an aggregate of two factors – the solution's price point and the value delivered. Let us discuss both factors.

1. Price and hidden costs: Here are some questions to consider.  

2. What are the implementation price and timelines?  

3. What is the cost of ongoing support?  

4. How does the price increase year on year?  

5. Are there any additional costs associated with Actions, integrations or roles?  

6. Are there any volume-based limitations? What happens when we cross that volume?  

7. What needs to be covered in the list price?

Value delivered

This is a trickier topic to evaluate when procuring an enterprise wide GenAI.

The value created can be qualitative and quantitative – both must be considered. Customer/employee digital experience or CSAT might not be equated in dollars readily – however, it might be a critical decision criterion for many enterprises. Regarding qualitative metrics, most product vendors will claim upwards of 70% reduction in cost or something equally impressive.

Understanding the cost-reduction aggregation

Primarily, cost-reduction for a company is an aggregate of multiple factors. The following questions can help you weigh these against multiple GenAI solutions.

• Potential for automation: What percentage of user requests may be resolved using GenAI? Although many vendors will put a large number here, seeking a breakdown of numbers would be great. For example, in the case of the IT Service desk, an effective GenAI system can resolve more than 65% of the tickets on its own – freeing the agents for more strategic tasks. You can checkout KPIs of a GenAI service desk here to map your own benchmark.  
• The ability to provide great answers: from knowledge will help reduce 10 to 20% of effort.  
• Automatically creating Smart tickets: using genAI will save another 10% of the effort.  
• Task and process automation: to save another 20 to 30% of manual human effort.  
• Desktop automation: to save up to another 15 to 20% of the human effort.  
• GenAI email responses: might save another 10% of the effort.

And so on….

Getting the most fitting break-up of numbers will help you understand the real potential of the solution. If your major consideration is IT, GenAI can be an incredible asset. You can see our webinar featuring Forrester here.

Consider the adoption rate of the GenAI product

There are meager chances of a 100% adoption of any product. There is always a percentage of users who will not use the product. And a healthy percentage of employee demographics is always keen to use an intelligent product. It would be wise to ask the vendor for the adoption rate of their product.

Our view: For your internal calculations, assume an adoption rate of 40 to 70% for a well-functioning product. Multiplying 'Potential for automation' with 'adoption rate' shall provide a realistic qualitative value delivered number. Remember to factor in qualitative value generated for the overall decision.

Rezolve.ai prides itself on providing GenAI solutions at a very affordable price-point, with demonstrable positive ROI from year 1.  

What does the product roadmap look like?

GenAI is a fast-moving technology; therefore, a product roadmap is an essential piece of information when making a decision. Many vendors will be willing to share this information at the right sales stage. Checking the roadmap will allow you to see if the vendor's overall vision aligns with your vision.

Support and service

Once the product is ready to go live, ongoing support becomes essential. Asking the following questions will come in handy when making a purchase decision for your GenAI solution.

1. What kind of support should you expect in case of product issues, outages or incidents?  

2. What kind of product training will be available for new team members?  

3. What kind of user resources are available?  

4. How difficult will reaching out to the customer success team or critical executive be?  

5. What kind of assistance is available for creating more automation?  

6. Is there a partner ecosystem to support the expansion of the use case?

Our view: Rezolve.ai is invested in customer outcomes and aligns its delivery to that. We believe subscribers should easily get value out of their investment.

We recommend you use this guide as a handy reference to evaluate potential Generative AI solutions for your company.

If you wish, you can always connect to the GenAI experts at Rezolve.ai and book a no-obligation call for more clarity.
‍

‍Don't Miss Out on Streamlining Your IT processes. Automate it with Rezolve.ai

‍
FAQs

1. What makes a GenAI solution right for my organization?
   Assess if the GenAI solution aligns with your specific use cases like knowledge management, productivity tools, or customer service.‍
   ‍
2. How do I choose between a single platform or multiple solutions for different use cases?
   Consider the specialization and functionality needs of your use cases. Some might be covered by a single product, while others might require specialized solutions.
   ‍
3. ‍Should I wait for existing products to incorporate GenAI features?
   Evaluate the GenAI roadmap of current vendors but also explore emerging vendors with a strong focus on GenAI for quicker adoption.
   ‍
4. ‍What is the importance of GenAI architecture in selecting a solution?
   The architecture, such as RAG-based solutions, significantly impacts the outcomes and effectiveness of the GenAI solution.
   ‍
5. ‍How should I approach the pilot project for a GenAI solution?
   A pilot project is recommended to test the solution's value and fit for your organization before full-scale implementation.
   ‍
6. ‍What are the key security and privacy considerations for a GenAI solution?
   Ensure the solution offers robust data security, privacy measures, and the ability to control PII exposure.
   ‍‍
7. Can GenAI solutions take actions beyond providing answers or summaries?
   Look for solutions capable of actionable support, such as ticket creation or workflow automation, with integration capabilities.
   ‍‍
8. What factors contribute to the ROI of a GenAI solution?
   Consider both the solution's cost and the value it delivers in terms of automation potential, effort reduction, and user adoption rate.
   ‍‍
9. How do I evaluate a GenAI solution's support and service?
   Assess the available support, training resources, and ease of access to the customer success team or critical executive support.