Generate Approved Business Content, With Responsible AI
Avoid Irrelevant, Perhaps Too-creative Business Documents
There has been a great deal of industry attention paid to creating content with Generative AI options. The public discussion of benefits and concerns has been quite broad: creating college essays, marketing content, marketing emails, television script copy and dialogue, but also legal documents. The consensus for most of the attention has been that Generative AI solutions can help by sourcing ‘starter’ content from the internet, with surprisingly creative and professional/human-sounding copy that can be generated from scratch, from simple prompts.
The business world is just starting to work through the pros and cons of this, given the many instances of ‘hallucinations’, falsehoods, copyright infringement lawsuits, plagiarism, and other potentially major issues. Businesses need to take a step back and carefully evaluate how this exciting and (at times) scary technology should be properly used.
Business Document Use Cases and Requirements
The business world uses many different types of documents that serve various purposes and have distinct requirements from one another. They are typically quite different from the above examples. Think about these specialized document types:
- Commercial property leases
- Commercial insurance policies
- Parts supplier contracts
- Software licenses or subscription agreements
- Negotiated master service agreements or statements of work
- Manufacturing product supplier contracts
- Privacy policies
- Employment agreements
- Proposals – responses to an RFP/RFI
- Distribution agreements
- Purchase or sale agreements
- Partnership agreements
- Clinical trial documentation
- Required public disclosure documents
- Non-disclosure / confidentiality agreements
- and many more
Business documents like these vary widely from one document type to another. There are nearly infinite variations of structure, language and content within any given document type. But they do have some major, broad characteristics in common:
- They are often authored or supervised by legal counsel, due to their inherent legal commitments.
- Each organization may have their own preferences for terms, language, and provisions.
- The resulting content and data is essential to the parties, so that they can properly execute them, follow obligations, comply with legal, operational and financial terms, and understand any risks.
- Industry standards or template language may come into play.
- They may be negotiated between parties, with varied commitments, and likely with deviations from the organization’s most common preferences.
- Marketing copy must not over-promise, so the organization does not under-deliver against those statements.
- Past agreements can be a guide to future agreements in most respects, because past agreements have been approved by all the internal stakeholders involved.
In short, there are a number of important constraints to pure creativity that apply to most business documents and agreements.
Traditional Workflow For Creating Business Documents:
Expensive, Expert, Personal Time and Attention Required
Today the common workflow to get important business documents done may look something like this:
- A designated business author combs through document repositories to find a starter document, or a relevant or required template document.
- The author makes a copy of the starter document and renames it.
- They update all the fields in the starter document or template, for the new names, dates, amounts, etc.
- They spend hours going through other documents for relevant clauses that may be needed for this particular case, but are not included or are different from the starter document. When found, these are copied and pasted, and likely edited, in the new document.
- The legal team may now need to provide legal review, to look for consistency, or compliance problems, risks, etc.
- Workflow tools may be applied to route the document to internal stakeholders.
- Documents are then sent to other parties for review or negotiation.
- Redline tools may be applied in the process to note editing or negotiated changes.
This can be a time-consuming, expensive process, to say the least.
Typical Generative AI tools do not speak to any of this. They simply provide blank-slate, starter content from sources outside the organization. While good for writing fiction, or college essays perhaps, this is not very helpful for writing legally-binding business documents.
Other recent tools attempt to take over the authoring process, and anticipate how you want a document to be generated. They take the ‘co-pilot’ approach where the author can figuratively step into the back of the plane while the AI takes over the flying, generating documents from the content and training it has access to.
Enter: 'Assisted Authoring' With Generative AI
There is another way to generate business documents, one that is more focused on the requirements of small and large enterprises alike. With an ‘assistive’ approach, the author or editor is always in charge, whether they are an attorney or business user. The technology exists to streamline the process by saving time, effort, and providing recommendations based on the organization’s previous best practices and current standards.
- Standard document repositories, workflow tools, and word processing tools can be used without buying into a whole new process.
- The generated recommendations are always there for the author to ignore, or use with a click.
- The recommendations can be content that the organization or legal team has previously approved, or forbidden.
It may sound like small differences: does the AI start content, or generate content for editing? Where do recommendations come from? The source and the nature of assistance is critical. Working with controlled, known recommendations keeps the author completely in charge of the intellectual work, while lessening the time burden and assisting with the quality control.
How Does Assisted Authoring with Generative AI Work?
By Highlighting your Best Practices
Docugami’s approach to assisted authoring begins by deconstructing your existing documents, focusing on each document type uniquely. Every word, phrase, clause, section, and subsection is indexed, so that they can be compared across documents of the same type. Commonalities, variances, and fields that are regularly changed are identified. Users can also flag important content for recommended use, or evaluation.
- The user begins, as usual, with a starter document in MS Word, and the redlining, editing, formatting and workflow is all totally familiar.
- When creating a new document, Docugami produces recommendations from this trove of sources and makes them available, on the fly, to the author via a MS Word add-in panel.
- Wherever the user clicks or hovers in the document text, seemingly by magic, recommendations for editing are highlighted, ranked, and made available at your fingertips.
- Terms that always change, are highlighted to the user. Clause recommendations appear as the AI detects variances across previous documents of that type.
It’s a lot like grammar or spelling correction tools, but goes further into the complete authoring process. The user can see all the proposed changes and decide whether to click - to choose, add, accept, remove, or reject content.
The result is that the user is working within your organization’s best practices, not incorporating new problems from who-knows-where. It does not attempt in any way to displace legal judgment, or creative judgment for that matter. But it does assist the attorney’s or business user’s work to make it much faster and easier, saving the organization time and money, and likely also avoiding missteps.