Skip to main content

​
Overview

Watch the video overview: https://youtu.be/IcQz4vlYaNE This guide covers the steps that apply to all ALK workflows, except the Reconciliation Workflow.
Use this guide to understand the standard steps for running any ALK workflow, including how to access a workflow, create sessions, upload documents, run processing, review results, and export outputs.

​
Before you start

Ensure you have the following ready before you begin:
  • Stable internet connection
  • Access to app.alk.work
  • Your client documents ready (follow the workflow’s allowed file types)

​
Access a workflow

1

Sign in

Go to app.alk.work and sign in with your credentials.
2

Open the Workflows tab

On the home page, select the Workflows tab on the navigation bar at the top.
3

Select a workflow

On the workflow page, scroll until you see Explore. Select the workflow you want to run. If you do not see it, click “See all”, then select the workflow.

​
Create a new template session

After selecting a workflow, you will see three steps for setting up your workflow:
1

Overview

This shows the workflow name, what ALK can do, the description or instruction provided to ALK, and the workflow use cases. Feel free to edit the workflow name and instructions as needed. If you are satisfied with the default template name and instructions, click Continue. If you make any changes, Continue is replaced by Reset and AI Assist. Click AI Assist to continue to the fields and allow ALK to generate the necessary fields based on your updated instructions. Click Reset if you want to revert to the default template.If you selected AI Assist, you will be asked to confirm whether you want to replace the default fields with the newly generated ones from ALK based on your updated instructions.
2

Fields

Here you will see the fields from either the default template or those newly generated from your updated instructions. There are five columns to note: field, task type, field type, dependency, and instructions.The field column contains the set of fields ALK generates based on your instructions, such as invoice number, vendor name, and other key information you typically review for the workflow you are running.For each field, you will see the task type. Some fields are Extraction, where ALK pulls information directly from the document. Others may be Classification, Validation, Reasoning, Summarization, or even Generate, depending on what you need.You can also choose the field type, such as text, number, or date. For example, if you want ALK to extract when an invoice was created, set the field type to date.Another powerful column is dependencies. For later fields, you can tell ALK which earlier fields to consider before answering, which helps keep outputs consistent across the entire screening.In the instructions column, you can fine-tune exactly what ALK should look for. For example, you might specify: “Extract the vendor’s tax identification number (TIN, EIN, VAT number, or equivalent).” These kinds of instructions help prevent common extraction mistakes and keep your results clean.You can also decide to add more fields where necessary.
3

Stages

Stages are the named stops in your workflow. They help you structure review and routing. They are essentially categories. You can create a stage that requires reviewer approval before moving forward.
To route properly to a stage, ensure you have a Recommendation, Decision, or Summary field set up. You can go back and change this field to a Single Select or Multi select field type, then route it to one or more stages. Also ensure the instruction given to ALK clearly states what to return and where to route it. For example, a stage instruction in this case could be: “Using PO Number and Line Items, determine whether a goods receipt or delivery confirmation is present and matches the invoiced items. Return ‘Matched’ if receipt evidence aligns, ‘Partial’ if only some items are confirmed, or ‘No Receipt’ if no receipt documentation is found.”
You can also include approvals for each stage where necessary. You can create stages such as Pending Approval, Approved, and Paid. Some stages are created by default depending on the template. In this case, once an invoice is matched or updated, you can route it to the Approval stage and assign a reviewer.If rejected, you can route it to another stage, such as Pending Approval, or create a Not Approved stage. This makes it easier to find, approve, and review invoices, documents, or entries by category or stage.After doing this, go ahead to start workflow.

​
Running a workflow

1

Start the workflow

After you click the Start workflow button, you will see a table where each field appears as a column.
2

Add documents

Click the add document button at the top left. Select one or more files from your device.Uploading multiple documents
  • You can upload up to 10 documents at once.
  • If prompted, choose how to upload:
    • Separate rows: For independent companies/documents.
    • Collection: For documents that belong together for one company. Name the collection when prompted.
3

Wait for upload and conversion

Wait for the upload and converting indicator to complete.
4

Confirm documents are ready

Confirm a new row appears for each successfully uploaded document. It will be marked “Ready” then you can go ahead to click run.

​
Run the session

1

Confirm documents are uploaded

Confirm the required documents have been uploaded.
2

Click Run

Click Run.
3

Wait for processing

Wait while the system processes the documents. Typical processing time is 1–2 minutes, depending on document size, quantity, and complexity.

​
Review results

  • Once processing is complete, extracted values will populate the table.
  • Confirm accuracy and completeness.

​
Viewing options

  • Table view (Default): Spreadsheet-style comparison.
  • Block view: Card-based view, easier to trace back to sources.
  • Wrap: Expands long text so it is not truncated.

​
Reviewing extracted fields

Switch to Block view to review and act on individual fields. Each field shows a status:
  • Verified (check icon): a reviewer has confirmed the value is correct.
  • Rejected (X icon): a reviewer has flagged the value as incorrect. Rejecting does not delete or change the value — it just marks it as not to be relied on.
  • Edited (pencil icon): a reviewer has manually changed the value.
  • No badge: the field shows a confidence indicator instead. Hover it to see ALK’s reasoning for the extracted value.
Available actions per field (if your role allows editing):
  • Verify / Un-verify: click the verify control to confirm a value, or click it again to undo. Un-verifying restores the field to Edited if it was ever hand-corrected, or to its original extracted state otherwise.
  • Reject: flags the field as incorrect.
  • Edit: click the pencil icon to open a text box and overwrite the value directly.
  • Retry: re-runs extraction for that specific field.

​
Export results

1

Click Export

Click the Export button at the top right and it exports to Analyst, where you can further edit the format or ask ALK for more information. Alternatively, you can directly export as CSV or Excel.
2

Select a format

Select a format:
  • CSV
  • Excel
3

Save and share

Save and share the exported file as needed.

​
Customise columns (properties)

​
How to customize a column

To edit the standard fields in any applicable workflow, follow these steps:
1

Open field options

Click the three dots next to a column to open the editing options.
2

Update property

Update the property name, type, task type, inputs, or prompt, then save.
3

Delete property

Remove the column entirely (only if it is redundant).

​
How to add custom columns

To add a new column (property) to your session:
1

Open the plus icon

On the far right of the table (the last column), click the + (plus) icon. This opens the Field options modal.
2

Fill in the property details

Fill in the property details:
  • Property name
  • Field type
  • Task type
  • Inputs
  • Prompt
3

Save

Click Save.
4

Run the session

The new property will be added to the table, and you can click Run to process your session.

​
Editing tips

  • Field type: This determines what the field (column) will contain, and how ALK will format the value it fills in when it runs your prompt (or extraction rules). For example, if you select Text, the result will be filled as text based on the prompt. Change this only if you understand the downstream effects. Options: Text, Number, Date, Single select, Multi select, URL.
    • Text: The result will be filled as normal text (words, sentences, short notes) based on the prompt. Use for names, IDs, comments, and any value that isn’t strictly numeric or date-based.
    • Number: The result will be filled as a numeric value based on the prompt, so it can be used for calculations, sorting, and comparisons. Examples: amounts, counts, percentages.
    • Date: The result will be filled as a date value based on the prompt (and time too, where applicable). Use for due dates, closing dates, or document dates.
    • Single select: The result will be filled as one option from a predefined list based on the prompt. Examples: Status, Risk level, Document type.
    • Multi select: The result will be filled as one or more options from a predefined list based on the prompt. Examples: Tags, applicable clauses, reasons.
    • URL: The result will be filled as a clickable link based on the prompt. Use when you want a source link, portal link, or any web link.
  • Task type: Determines system behavior:
    • Extraction: Pulls an exact value from the uploaded file.
    • Classification: Assigns a value to a category.
    • Validation: Checks for completeness or format.
    • Reasoning: Evaluates or derives a value based on inputs.
    • Summarization: Condenses text into concise notes.
  • Inputs: Select the other columns you want ALK to consider when producing an output for this specific field. These selected fields act as the reference context or supporting data ALK can use when following the prompt (especially for reasoning fields).
  • Prompt: Keep guidance clear and specific.

​
Troubleshooting

  • If a session does not run, confirm documents are fully uploaded and retry.
  • If results look incomplete, confirm the document is readable and contains the needed information.

​
Need help?

If you run into issues, contact the Wamiri support team at support@wamiri.com. Include:
  • What you were trying to do
  • What happened instead
  • Any error messages
  • Steps to reproduce
  • Screenshots or recordings (if available)