How To Maximize Value
It’s very easy to misuse Documents Intelligence (DI) and have the wrong first impression.
The last thing we want is for you to dismiss DI as incapable because you tried doing something that it was not designed for.
Understand Tenants
First, you need to understand what kind of information each tenant can give you:
-
Presales Archive - you can find clients who had problems similar to the one you are working on, you can find solutions we proposed in response to those problems. Not all of these solutions were actually used and deployed. We win less than 50% deals that we bid on. The
presales archivetenant exposes you to the 100%. -
Production Projects and Contracts. You may think that an SOW equals a project. You would be correct for a good number of projects and wrong for many more. We have umbrella contracts (where one SOW triggers tens if not more individual projects each pursuing different goals, using different technologies, and solving different problems). We also have umbrella projects (where one project code runs through multiple SOWs each focusing on something very different than the others). So it is many-to-many in the EPAM world between projects and contracts. The
production projectstenant has project at the center and thecontractstenant has the SOW at the center. The other thing that is important to note is that:
We read people’s experience as the main content to understand projects. We supplement that with the SOW scope summary where available but the primary goal is to see the project from the eyes of people who participated in it. We read the SOW and its amendments as the main content to understand contracts. The two are similar but not the same.
- Professionals includes active EPAM employees across all tracks (A, B, C, O) and levels (1-5). We filter out non-delivery roles such as Students, Administration, Legal, Staffing, Service Desk, Talent Acquisition, People Services, Finance, People Management, and Talent Development Services. The focus is on professionals actively involved in project delivery and client-facing work. Go to Telescope for all employees. We are not trying to replace it. Our goal is to surface professionals with relevant experience when you ask DI for information.
If you know a person X is an expert in something and in general a great professional but you don’t see any of it in DI, ask that person X to please update his or her profile with their past project experience, as well as their skills and contributions. We can’t read people minds so we have to go based on what we find in our internal systems.
Use Browse
When your question is of the Talk To Your Data type, you need to browse, not search. For example, if you want to find experts who know AWS, write node.js, and have worked with Amazon Connect, you should not try to search for it in the conversational interface. You should instead go to the Professionals tenant and use the facets on the left of your screen to formulate this search. This will be much more precise.
If you can rephrase your question as a SQL query (aka talk-to-your-data), you should browse, not search. We
will build agents into our agentic conversational search that understand these types of questions and do the
query-based search for you, but for now you need to do it yourself
When you browse, there is a search bar at the top of your screen. This one is not a conversational search. Do not write natural language english sentences into it. You will get garbage out. This is your very common keyword search - find a person by name, project by code or a client name, contract by client or title, etc.
Use Search
What do you do today when you need to find if we have ever done X for another client? Or when you wonder how to approach a problem Y and if we have any relevant case studies? I will tell you what you do - you either ask someone you know or you write an email to WFT Project Managers. How do you write that email? You give some context. You explain where you or your client is coming from. You explain the constraints that apply in your case. You ask for specific expertise in various technologies or business domains or problem areas or all of the above.
This is how you use search.
To get the most out of DI, your search chat message is your email to WFT PM. You give context, you spell out
constraints, you reference relevant technologies and business areas and problem spaces. You spend five minutes of your
time to type or dictate your question. If you give it five words instead of five minutes, we call it a stupid question and we don’t feel bad when DI cannot be very helpful. Help us help you
Use Collections
Last but not least, do not stop at asking the question and admiring the answer. Read into it and pin what’s relevant to what you are looking for to a collection. Then search again using a different combination of words and keywords. DI is LLM-first and that means even the same question asked twice is likely to yield different results. Don’t expect DI to just give you exactly what you are looking for even if you have spent the five minutes we are asking you to spend to formulate your ask.
If you have additional materials that support your inquiry, add them to the collection and then ask detailed questions with the AI focused on what you deemed relevant. You can drill into the results, you can ask specific questions about an entity you added to the collection. Type @ and we will give you a list of all the things you added to the collection so that you could tell AI which of the entities you want it to zero in.
Think
DI is not a magic wand. DI does not read your mind. DI can be a very helpful tool if you know how to wield it and if you are willing to do your part. The thing is, if we could answer your question precisely and fully, why would we need you? Think about it every time you evaluate an AI application. You need to realize that the help you get from AI is amazing and better than having no help even if it doesn’t fully replace your effort. If/when it does, it will replace you…