| Criterion | Weighted Multiplier |
Details | Score 0–2 |
Notes / Evidence |
|---|---|---|---|---|
|
Fit to our documents
|
Can it handle our most used documents?
|
|||
|
Accuracy on key fields
|
Is it accurate on what we care about? Can we see the
exceptions?
|
|||
|
Unstructured Capabilities
|
Is it understanding our most complicated documents?
|
|||
|
Speed & capacity
|
Can it handle large volumes at speed?
|
|||
|
Machine Learning
|
Does the solution exhibit continuous learning based on past
corrections?
|
|||
|
Human-in-the-loop
|
Can a human easily review mistakes and make changes to the
model training?
|
|||
|
Cost clarity
|
Is it outcome-based? Token-based? Usage-based? Are we okay
with that?
|
|||
|
Deployment
|
Can we choose where it runs based on our security and
compliance needs?
|
|||
|
Auditability
|
Can we control how the model is trained, monitor changes, and
see how our data is used?
|
|||
|
External users
|
Can external users submit documents and are they notified when
inputs are missing?
|
|||
|
Integration
|
Does it support the neccessary output formats to integrate
results in our apps and databases?
|
|||
|
Vendor focus
|
Does the vendor have a clear vision for the future with a
product roadmap?
|
Return as many times as necessary to narrow-in on the best solution for your needs!
For more collaboration or consultation, book a meeting with our team of engineers.