AI and Rules - Recent Threads

Chat Agent – Tool Execution Failure ("Downstream tool request failed")

Kaviarasu Mohan — Thu, 09 Jul 2026 11:00:15 GMT

Hi Team,

Our chat agent is intermittently failing when attempting to execute tool calls. End users are receiving one of the following error messages during these interactions:

Tool execution failed: Downstream tool request failed
I'm currently experiencing a system issue retrieving the data.

AI Agent Performance

Mallepu Prashanth — Thu, 09 Jul 2026 10:16:13 GMT

How accurate AI Agent works same task with different data. Will it hallucinate or imagine things and provide output? Does it reconcile based on the repeated tasks or does it take every new data as new task?

Getting an error in process model: Agent 250512b0-a322-45b6-85c5-525b87755a8c could not be found. AI Agent is not working in Process Model.

Mallepu Prashanth — Thu, 09 Jul 2026 10:03:56 GMT

When I test AI Agent it is working as expected and giving result. But when I incorporated it within Process model, it is giving error: Agent 250512b0-a322-45b6-85c5-525b87755a8c could not be found. So, I created another AI agent and getting same error stating Agent could not found. I have double checked the UUID of the AI Agent it is correct. What could be the reason?

In Admin console for our environment AI services are enabled and active.

AI Doc Center vs AI Skills

kl0001 — Thu, 09 Jul 2026 01:28:25 GMT

My use case is that I want to be classify documents into one of several categories, then depending on the category, extract certain fields from it. My question is for this kind of use case, what exactly is the difference between using AI Skills to do this versus using AI Doc Center? Looking around, I see some people saying AI Doc Center uses more tokens and that it can handle more complex cases, but I can't see anything in documentation about if that's actually true or not, and anything about when you'd want to use one or the other. I've been in loops reading the same page over and over again, so at this point I might have missed something, but if someone's able to help point me towards an answer, that would be most appreciated!

a!agentChatField not rendering/connecting to Chat Agent — need help

Kaviarasu Mohan — Wed, 08 Jul 2026 04:33:16 GMT

I have configured a Chat Agent named "Roster Assistant" using a!agentChatField with the following setup:

The agent is connected to our Roster Record Type as a tool.
Field-level instructions have been configured within the record type.
Agent goals and instructions are designed to:
- Respond to greetings such as "Hi" or "Hello".
- Retrieve and return roster information for users.

Problem

The agent does not respond as expected. Regardless of the message sent—including simple greetings like "Hi"—the chat either takes an unusually long time to process the request or eventually returns an error.

Additionally, when reviewing the execution in the Monitor tab, the agent run remains in a "Processing" state for an extended period before ultimately failing with an error.

As a result, the agent is unable to provide responses for even basic interactions, making it difficult to determine whether the issue is related to the agent configuration, record type tool setup, or underlying processing behavior.

MCP Server endpoint URL in Appian 26.6

Raúl Gómez Moya — Tue, 07 Jul 2026 19:24:15 GMT

Hello everyone,

I’m configuring the Appian MCP Server in Appian 26.6 and I noticed a difference between the endpoint URL shown in the Administration Console and the one described in the documentation.

In the Administration Console > MCP Server, the endpoint URL is displayed as:

<ENVIRONMENT_URL>/mcp

See Image 1.

However, the documentation for enabling the Appian MCP Server says that the endpoint URL follows this format:

<ENVIRONMENT_URL>/suite/mcp

docs.appian.com/.../enable-configure-mcp-server.html

See Image 2.

My question is:

Which endpoint should be used when configuring an external MCP client in Appian 26.6?

Should it be:

<ENVIRONMENT_URL>/mcp

or:

<ENVIRONMENT_URL>/suite/mcp

Has anyone configured this successfully and can confirm which URL is correct?

Thanks in advance.

Attachments

Image 1: Administration Console showing the MCP endpoint as /mcp.
Image 2: Documentation showing the MCP endpoint format as /suite/mcp.

How to Check AI Token Usage for AI Skills in Appian?

sureshs276515 — Tue, 07 Jul 2026 14:18:31 GMT

Hi Team,

We are evaluating AI Skills/Generative AI usage in Appian and would like to monitor token consumption for optimization and cost governance purposes.

I have a few questions:

Is there a way to view the input tokens, output tokens, and total tokens consumed by an AI Skill execution?
Does Appian expose token usage metrics at the AI Skill or process level?
What is the recommended approach to track and analyze token usage for prompt optimization?
Has anyone built a custom monitoring/dashboard solution to estimate token consumption for AI Skills and Agents?

thanks in advance.

AI user chatbot

rohity0003 — Mon, 22 Jun 2026 15:09:40 GMT

Hello,

I have built a chatbot in appian. I have two objects in my interface. First is a text box, which takes user question. And, the second is a button user clicks to send question to AI bot and get response.

Currently, when text box loses focus, a local variable called isProcessing is turned to true(So the user can be shown that processing is occurring). And, when the button-press returns with AI response, the isProcessing is turned to false, and the AI response is shown to user.

I want to have it so that when the text box loses focus, the local variable isProcessing is turned to true, as well as data being sent to AI bot. And, then when AI bot returns, the isProcessing is turned to false, and result shown.

Is this possible?

Thanks for your time.

Data Fabric Chatbot — Actionable Interface

Kaviarasu Mohan — Wed, 17 Jun 2026 14:15:09 GMT

I'm currently working on a case management project in Appian and exploring the potential of the Data Fabric Chatbot as a conversational interface for employees to manage their work. While the chatbot works great for querying and retrieving data. But today, I'm wondering if Appian has any plans to evolve it into a fully actionable interface in future releases.

User: "Give me all my open cases"
Chatbot: Returns a list of open cases assigned to the user in the table.

User: "Open the high priority case"
Chatbot: Opens the case in a side panel or new screen

User: Uploads a document, fills in the required details, and submits
Chatbot: Confirms the case is completed

This would basically make the Data Fabric Chatbot a one-stop shop for employees — no more jumping between screens just to find and complete a case. Everything handled right there in the chat.

AI Skill

kunalj445796 — Tue, 16 Jun 2026 06:22:04 GMT

Hello All,

I am using Version 26.3 and while additing new AI Skill, i am getting blank screen everytime. I cleared cache, restarted browser and system but with no luck. Even tried with other bowsers also. Within my organization, the same is working fine for few but there are few who are facing issue similar to what i have been facing. Please suggest.

Br,

Kunal

How to search values within constans

Dusty — Tue, 09 Jun 2026 22:04:34 GMT

Hi All,

I have a use case where we have to search for values in numerous constants if it matches a specific email domain. Just applying the search box value isnt productive as it doesnt provide the results with actual constant values, also gives descriptions which is not useful in my case.

How to Process Multiple Invoices Using Document Extraction?

mohitr083133 — Thu, 04 Jun 2026 06:31:41 GMT

Hi Everyone,

I'm using Appian Document Extraction and need to process multiple invoice documents uploaded in a claim.

What is the recommended approach to:

Pass multiple invoices for extraction
Extract data from each invoice
Store the results in a List of Record Type/CDT

Can a list of document IDs be passed directly to Document Extraction, or should each document be processed individually (using a loop/subprocess)?

Queries on AI Doc center classification

appalasurip0001 — Thu, 28 May 2026 19:54:55 GMT

Hi,

I am working on DocCenter Classification and have a few questions around reconciliation and automation. Looking for guidance from anyone who has implemented this.

**My Use Case:**

I want to classify documents automatically without any human involvement in the reconciliation step.

**My Questions:**

1. Is it possible to auto-reconcile a classification instance in DocCenter without human intervention?

2. Does DocCenter Classification expose a confidence score as an output in a process model (similar to the Classify Documents Smart Service)?

3. If DocCenter Classification does not support auto-reconciliation natively, what is the recommended approach to achieve straight-through processing (STP) for classification?

Thanks in advance.

Appian GenAI Excel Classification Issue with Data vs Empty File

sureshs276515 — Tue, 26 May 2026 18:27:35 GMT

Hi everyone,

I’m facing an issue while using Appian AI Skill (GenAI) with an Excel file.

When I upload an Excel file with empty data, I get one type of classification output.
However, when I upload the same file structure but with data inside, it produces a different classification output.
Essentially, the classification result varies depending on whether the file has data, and it seems inconsistent each time.

Has anyone experienced this issue before? Any insights on why the classification behavior changes like this?

Inconsistent data types (Text vs Variant) in List of Map output

rajeshp7374 — Tue, 26 May 2026 13:51:30 GMT

I am working with a rule in Appian where I expect a consistent structure of List of Map, but I am seeing inconsistent data types in the output.

a!localVariables(
local!indexval: index(
ri!userFilters,
"userFilter",
null
),
local!filters: split(local!indexval, ";"),
a!forEach(
items: local!filters,
expression: a!localVariables(
local!valueStart: find("value=[", fv!item),
local!valueBlock: if(
isnull(local!valueStart),
"",
mid(
fv!item,
local!valueStart + 7,
len(fv!item)
)
),
local!cleanValueBlock: substitute(local!valueBlock, "]", ""),
/*Values */
local!AccountName: if(
find("name=AccountName", fv!item) > 0,
a!forEach(
items: split(local!cleanValueBlock, ","),
expression: trim(substitute(fv!item, """", ""))
),
{}
),
local!ProductCategory: if(
find("name=Product Category", fv!item) > 0,
a!forEach(
items: split(local!cleanValueBlock, ","),
expression: trim(substitute(fv!item, """", ""))
),
{}
),
a!map(
AccountName: local!AccountName,
ProductCategory: local!ProductCategory
)
)
)
)

expecting ProductCategory and Account name in single map

Ai Agend having issue

pradeepkumawat — Tue, 26 May 2026 08:52:42 GMT

I am exploring the Appian AI agend and faced the below Appian error. enable to understand and fix these issue.
If anyone has faced similar issue please let know the fixes.

LLM execution failed: EVENT_NOT_FOUND - No event exists with name: setVariableValue

pradeepkumawat — Wed, 20 May 2026 08:26:03 GMT

I am facing error in the AI agend while reading the document.
LLM execution failed: EVENT_NOT_FOUND - No event exists with name: setVariableValue
Steps1: When I am passing this document as in the TOOL in the AI agend than it is working fine.
Steps2: When passing this document in the input than facing this issue.

Please just How can I pass document dynamically in the process model?

AI Document Extraction

Kavya — Sat, 16 May 2026 06:33:20 GMT

Hi Everyone,

We have a requirement where we are using AI-based document extraction to extract fields from a predefined form.

In the same form, if there are any empty spaces and the user accidentally adds handwritten text or notes outside the expected form fields, we need to detect that additional handwritten content separately.

Is it possible in Appian with help of AI extraction to achieve this use case ?

Kind Regards,

Kavya

Smart Search Indexing Failure

jamesm4933 — Thu, 30 Apr 2026 16:20:13 GMT

I have a Record Type with a Document record field. We have a business requirement to be able to use Smart/Semantic search on this and a text field.

The Smart Search Sync consistently fails with the following error message, and a list of failed record ids: "Cannot retrieve the document. Verify the document management record type is configured correctly."

Through investigation, we discovered, that the Document field for these records was NULL, which is consistent with the above error message.

We would like to have a successful search indexing, for this record type, and so we removed those records from the dataset temporarily. However, now we get the following error message: "We couldn't check the smart search status. Try resyncing your record type."

Attempting to Sync is what produced this error, so it seems a bit unhelpful as an error message.

Why is the indexing failing when the selected smart search field is empty? Is there a workaround?

Now that the Smart Search syncing is failing with no further information, what can we do?

I am using a!currency to show thousand separator

shridevim9458 — Fri, 17 Apr 2026 10:12:45 GMT

I am using a!currency to show thousand separator but it by default shows $ symbol if used USD ISO.Here I need to remove this $ symbol while displaying this field.

AI Agent: INVALID_TOOL_RESPONSE

David G. — Wed, 15 Apr 2026 19:19:59 GMT

Sometimes, when I test an agent, it processes for a while, then after a few minutes throws this error:

System Error

This appears to be a system issue. Try again later or submit a support case if the problem continues.

Source: Run Agent - Agent: DG_testAgent - Run ID: 52

Details: LLM execution failed: INVALID_TOOL_RESPONSE - Required schema field not provided in data tree: textValue

Any suggestions on what could be going on? Sometimes the agent succeeds, but the frequency of getting this error makes me concerned about using this in production.

26.3 environment hosted on Appian Cloud.

"read msg" smart service with .msg to html confusing to AI in some instances

ericc087568 — Fri, 10 Apr 2026 21:36:22 GMT

I have a subset of .msg files which when read with the read msg smart service, render in a tiny font with the .html output which results in AI being unable to interpret.

Would I probably be better off running several reg ex statements to exclude tags that are confusing it and pasting in some style that would force the font, or just use the extracted (not the generated result document), or move it to PDF, or some different approach?

Regards,

Eric Casteel

AI Doc Center

shivakanthreddyp0001 — Tue, 07 Apr 2026 00:05:48 GMT

Hi All,

We were previously using AI Document Center 2.9, and whenever we created a classification model, the system would automatically create the corresponding classification AI skills in the Generated Objects application.

Recently, we upgraded to Document Center 3.1. After the upgrade, when we create a classification model, the AI skill is no longer created automatically.

Upon reviewing the Create Classification process model, I noticed that it was updated as part of the upgrade and no longer calls the “create objects” process model, which the earlier version did.

Is this the expected behavior in Doc Center 3.1?
Or is there any additional configuration or post‑upgrade step that we might have missed to enable automatic AI skill creation?

Thanks in advance

AI Skill Doc Extraction

leelaraok0002 — Fri, 27 Mar 2026 16:24:58 GMT

can we configure human task where api kicks appian process which ahs doc extraction ai skill in case doc extraction fails

What is a best practice for checking extraction instance result in AI Doc Center for further routing to custom reconcile?

ericc087568 — Thu, 26 Mar 2026 22:06:39 GMT

First of all, AI Doc Center is fantastic (3.1.3 installed). Love the way it is easy to set up and version classification and extraction, so much right out of the box.

I think I'm missing something in the documentation though on AIA Extraction Instance's status in a 'real' extraction. I see some constants in the AIA app and can infer in a different app that a status of 2 or 4 means a particular thing that might send me to my custom interface for reconciliation of the instance, which in turn would trigger a model to save to data fabric if reconciled and not abandoned / canceled by an end user. But, I'm getting off track...

I guess I'm asking what is the best practice to see the AIA Extraction Instance's notion that accuracy / precision / validation rule fail on an extraction would drive the decision to send a custom user reconcile task. At the moment, I would be cobbling together a big 'Or' to see if any reconciliation should be routed.

Regards, and see anyone at Appian World 2026?