Product Gap

Smart Templates (AI-supported real-time format support)

Impact:

Small

Project Size:

Medium

What

System for helping ensure that the outputs of a task are in the right form, e.g., by evaluating content’s fit to a specified template and instruction, and giving feedback as comments, suggested edits, or through chat.

Why

Increase sophistication of participant flows ↑, efficient facilitation ↑, more substantive outputs ↑ and more efficient ↑ processes.

Problem Definition

Many processes with highly structured outputs (e.g. citizens’ assemblies) are sequences of exercises that rely on participants completing specific templates whose outputs feed into subsequent exercises. Without close facilitation, sometimes the outputs may not be structured correctly or in an actionable form; this level of facilitation micromanagement is resource-intensive.

Definition of Success

Automated support provided to participants is equivalent to or better than close table facilitation and is provided at little to no cost with reduced bias. For a Citizens’ Assembly (100-people, ~20 table facilitators, 2 days of table facilitation for relevant exercises), we estimate a saving of ~$40,000 in reduced human workload. This is often financially infeasible and so not done in many processes, this would significantly raise their quality.

Requirements

Real-time feedback and support for template adherence.
Two kinds of interface: within collaborative writing software (e.g., Google Docs extension); and interface for getting feedback on physical templates (e.g., uploading photos).
Very easy to use for participants with minimal device experience: web and mobile; auto focus; accessibility options; and clear instructions.
Group work capabilities: parallel work; document locking; and state-save.
Guarantees around bias.

Existing Limitations

This is either done manually with facilitators checking task completion, or done in an ad-hoc, unintegrated fashion with AI tools (without ideal workflows, transparency, bias guarantees, etc., and which still requires supervision). In some cases participants are in open dialogue with GPT’s (with no specific task context, but with process information in knowledge). Flaws include participants being overly swayed by verbose writing, allowing GPTs to rewrite content in a way that could change meaning, and losing track of task focus.