When your role requires you to generate a high volume of documents, particularly when those documents follow familiar formats, that’s a great opportunity to have AI assist in the process. Its simply faster and more efficient to use intelligent automation to do the leg work.
Document completion tools fall into two categories – simple templates for basic document creation and composite templates which synthesize information from a variety of sources and combine them into one finished product.
Single documents may include:
- Corporate profiles.
- Financial Spreadsheeting
- Case summaries
- Biographies
- Session summaries
Composite documents may include:
- RFP (request for pitch) responses
- Credit memos
- Pitchbooks
- Model documentation
- Analyst reports
- Sales team scripts
These are just a few examples of the type of documents that AI-powered document completion systems can handle. They are hugely flexible, however, and we’re discovering new use cases all the time.
Three ways to use document completion tools
As well as generating different kinds of documentation, both simple and composite, AI can pull data from multiple sources, with different data sources, targeting methods, and prompts. This mitigates against any “mission creep” and ensures the generative AI draws the information you need with accuracy and specificity.
There are three main ways to prompt AI for document creation, whilst restricting the AI’s effects to just the intended purpose:
1. Standardized Prompts
Where it is important a document follows a particular format, and draws from similar sources, standardized prompts can be used. This ensures the resulting documents are readily comparable, follow a recognized format, and that important information is foregrounded.
Example: Creating company profiles for investment banking pitchbooks.
2. Selective Inputs
Many financial documents, such as spreadsheets, require highly structured and accurate data, pulled from very specific sources. In such cases, it is important not all allow the AI research tool to roam freely in test data. Instead, it must be directed towards finite, pre-ratified inputs.
Example: Descriptions of financial data extracted using a deterministic tool like Spreadsmart. This is our digital platform to pull data from financial statements, and it has been finely tuned for that purpose.
3. Enterprise searches
Sometimes it is important to throw the net wide across a business or sector. This approach is necessary when producing detailed and complex documents that argue a particular business case. It is vital when producing documentation in a highly competitive tendering or bidding process.
Example: This approach works for generating a “draft zero” RFP response, obtaining data from a wide range of data sources, both structured and unstructured.
Evalueserve’s AI proposal solution generates high quality drafts using information from across a given enterprise including past proposals, case studies, and more. The initial draft is then checked and finalized by domain experts and other stakeholders prior to submission.
Case study: Generating a bid-winning RFP for a global consulting client
Optimizing an approach to competitive bidding within tight deadlines.
Challenge:
Our client, a global consulting firm, found it challenging to maintain consistency when producing more than 200 RFPs across six distinct business units annually. Their existing methods were highly resource intensive, and comparatively inefficient. To keep up with the competition such processes had to be streamlined and improved.
The consultancy wanted to optimize the process, automating where possible and quickly generating thorough and well-argued proposals highly likely to secure new business, in advance of their competition.
Solution:
We implemented Publishwise, our knowledge asset reuse and discovery platform, to streamline key workflows for our client. The client’s research processes were further enhanced using GenAI to create drafts for common proposal sections. Our tools’ GenAI capabilities were specifically configured to the client’s needs, in collaboration with domain experts.
For example, we’ve built a prompt library to help standardize outputs. This saves time and helps counter GenAI’s tendency to produce excessively diverse or unpredictable results.
During set up, we refined the client’s data corpus by integrating all essential proposal inputs and converting them into a vector database. The vector format functions as a digital filing system for large language models. It’s an essential tool for sifting through data sets at speed.
Adopting this approach often surfaces deeper and more contextually specific insights than using traditional search methods.
Result:
Our GenAI solution produced reliable analytics, pre-formatted for ease of interpretation, which allowed our client to trust the information they were receiving and achieve a 40% increase in response times.
By saving valuable preparation time, the client was able to focus more effort on contacting new leads, securing clients, and winning contracts.
