Generative AI-assisted EDI mapping
The AWS B2B Data Interchange generative AI-assisted EDI mapping capability expedites the process of writing and testing bi-directional EDI mappings, reducing the time, effort, and costs associated with migrating your EDI workloads to AWS. This capability leverages your existing EDI documents and transactional data samples to generate mapping code using generative AI. You can then use the generated mapping code as a starting point and further customize it to produce output formats that align with downstream data integration needs.
Topics
Prerequisites for using the AWS B2B Data Interchange generative AI-assisted EDI mapping capability
Before you can use this feature, you need to enable the models in Amazon Bedrock.
Note
You do not incur additional AWS B2B Data Interchange charges to generate mapping code beyond the standard Amazon Bedrock Pricing
To enable models in Amazon Bedrock
-
Sign in to the AWS Management Console and open the Amazon Bedrock console at https://console.aws.amazon.com/bedrock/
. -
From the left-hand navigation menu, choose Model access from the Amazon Bedrock configurations pane.
-
Enable all Anthropic models (AWS B2B Data Interchange currently uses Claude 3.5 Sonnet v1 and Claude 3 Sonnet), then choose Next.
In the future, there may be newer models that we will suggest that you enable as well.
-
If prompted, you may provide your use case details for access to the Anthropic models, including your company name, URL, industry, user persona, and description of your use case.
-
Review your selected models, and if you don't need to make any changes, choose Submit.
Notes about generative AI-assisted EDI mapping
Note the following:
-
This feature is available for both inbound and outbound EDI processing.
-
To use this feature, you must upload input and output samples when you configure a transformer.
-
An accuracy score is generated for each mapping, to help you determine whether additional edits are needed.
-
Mapping code is generated during configuration of your transformer resource, and not a transformation runtime.
-
No customer data is stored or used to train the models: each mapping generated is a one-time operation.
-
Currently, the AWS B2B Data Interchange generative AI-assisted EDI mapping capability is only supported in the US East (N. Virginia) and US West (Oregon) regions.
Using generative AI-assisted EDI mapping in AWS B2B Data Interchange
The transformer configuration wizard has three steps:
-
Transformer configuration
-
Mapping configuration
-
Review and create
For details on creating bi-directional transformers, see Create an inbound transformer or Create an outbound transformer.
To use the AWS B2B Data Interchange generative AI-assisted EDI mapping capability, make sure to upload both an input and output sample in the transformer configuration step (Step 1) when creating or updating your transformer resource. If you’ve specified both an input and output sample in Step 1, you see the Generate Mapping option enabled during the mapping configuration step (Step 2).
To use generative AI-assisted EDI mapping in AWS B2B Data Interchange
-
Upload your EDI document sample and JSON or XML data file sample to an Amazon S3 bucket (or buckets) with the appropriate policy and permissions. For details, see Amazon S3 bucket policies and permissions.
-
Navigate to the transformer homepage in the AWS B2B Data Interchange console. Choose Create transformer to create a new transformer or select an existing transformer from the list and choose Edit to update the configuration. In the Sample documents section, specify the input and output samples that you uploaded to Amazon S3 in the previous step.
-
Select Generate Mapping. You can view the progress and percentage complete in the progress bar.
In the Generate Mapping - optional pane, notice the following note: By selecting Generate Mapping, you acknowledge that additional charges will be incurred for the use of Amazon Bedrock.
The Mapping editor pane will be empty before you select Generate Mapping and while the mapping is being generated.
This screen shows the Mapping editor panes before you select Generate Mapping and the Mapping pane is empty.
-
When the mapping completes, you see the Generate Mapping was successful message. You also see the Mapping editor has been populated with mapping code.
After the mapping has been generated, the Diff and Accuracy details displays the Mapping accuracy score and the Mapping evaluation. Note the following:
-
The accuracy score remains active as you continue to make manual edits, and changes based on your editing.
-
The score is determined by how well the provided sample output matches against the output document that is generated by the generative AI-assisted EDI mapping.
-
The generated score is determined by counting the number of matching lines and divides by the total number of lines in the original output document. For example, if 19 of 20 lines match, the accuracy score is 95%.
-
If you are unsatisfied with the mapping, return to Step 1 to specify alternate input and output samples. Then, return to Step 2 and select Re-generate Mapping. Using alternate EDI document and JSON or XML data file samples will result in new mapping code.
-
-
When your mapping is in a satisfactory state, select Next to proceed to step 3, Review and create.
Continue to the Review and create step, as described in Create an inbound transformer or Create an outbound transformer.
