What challenges did Visma face before AI implementation?

Manual validation of complex travel and expense claims took up to 1 hour per multi-month claim. About 9.58% of claims were returned for correction, extending reimbursement times and adding workload for finance teams. Contextual free-text justification and Finnish tax or policy rules were difficult to automate with conventional rules-based logic.

What was the impact of the AI solution?

Visma helped customers cut average claim lead times by 45%. Claim returns were reduced by 25%, with some customers seeing reductions as high as 50%. Complex multi-month claims that once took up to 1 hour can now pass validation in minutes. Visma expects a 25% increase in total annual recurring revenue within 2 years, with potential to reach 30% as adoption expands.

Rank#653of 3,449

Source

ExploringEvidence: Medium65/100

Visma automates travel and expense claim validation with Amazon Bedrock

Use case typeClaims automationUpdated Jul 15, 2026

Visma, through M2 by Visma, built M2 Tarkka to automate validation of travel and expense claims on AWS. The solution combines deterministic rules with LLM-powered checks to interpret contextual free-text descriptions and apply Finnish tax and policy rules. Visma also added an AI agent so customers can upload travel policies and ask natural-language questions with source-attributed answers.

Organization: Visma
Industry: Finance
Location: Finland
Published: July 2026

Reported outcomes

−45%

average claim lead timesOther quantified impact

−25%claim returns−50%claim returns−87.5%claim review time+25%annual recurring revenue

Strategic outcomes

Employee experienceReimburses employees soonerCost efficiencyReduces manual finance workloadOther strategic outcomeAdded customer policy Q&A assistant

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Average claim lead times: 45% decrease

AWS Customer StoriesJul 15, 2026Customer storyExplicit claimMedium evidence strength

helped customers cut average claim lead times by 45 percent

Normalized claim

Claim returns: 25% decrease

AWS Customer StoriesJul 15, 2026Customer storyExplicit claimMedium evidence strength

reduce claim returns by 25 percent

Normalized claim

Claim returns: 50% decrease

AWS Customer StoriesJul 15, 2026Customer storyExplicit claimMedium evidence strength

some organizations are seeing reductions as high as 50 percent

Normalized claim

Claim review time: 87.5% decrease

AWS Customer StoriesJul 15, 2026Customer storyInferred claimMedium evidence strength

could take up to 1 hour ... now pass validation in minutes

Normalized claim

Annual recurring revenue: 25% increase

AWS Customer StoriesJul 15, 2026Customer storyExplicit claimMedium evidence strength

expects M2 Tarkka to drive a 25 percent increase in total annual recurring revenue

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Visma
Provider: AWS
Maturity: Exploring
Linked source: AWS Customer Stories

The module uses about 50 validation rules, combining deterministic checks with LLM-powered evaluation

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Claims automation
2Workflow automation
3Conversational assistants

Manual validation of complex travel and expense claims took up to 1 hour per multi-month claim.
About 9.58% of claims were returned for correction, extending reimbursement times and adding workload for finance teams.
Contextual free-text justification and Finnish tax or policy rules were difficult to automate with conventional rules-based logic.

Visma developed M2 Tarkka as a cloud-based microservice on AWS.
The module uses about 50 validation rules, combining deterministic checks with LLM-powered evaluation.
Amazon Bedrock is used to interpret contextual descriptions and apply nuanced tax and policy rules.
Amazon Bedrock Knowledge Bases power an AI agent that answers company-specific travel policy questions with source attribution.
AWS Lambda handles event-driven processing, while Amazon ECS and AWS Fargate run containerized workloads.

Technologies

Amazon Bedrock Amazon Bedrock Knowledge Bases AWS Lambda Amazon ECS AWS Fargate

Visma helped customers cut average claim lead times by 45%.
Claim returns were reduced by 25%, with some customers seeing reductions as high as 50%.
Complex multi-month claims that once took up to 1 hour can now pass validation in minutes.
Visma expects a 25% increase in total annual recurring revenue within 2 years, with potential to reach 30% as adoption expands.

Architecture

M2 Tarkka runs as Visma's first microservice built entirely on AWS. The solution combines approximately 50 validation rules, some deterministic and some powered by LLMs. Amazon Bedrock is used for model-assisted evaluation of contextual descriptions, Amazon Bedrock Knowledge Bases support policy Q&A with session context and source attribution, AWS Lambda handles event-driven processing, and Amazon ECS with AWS Fargate run containerized workloads.

Sources & evidence1

Evidence: Medium65/100Evidence strength