> ## Content Index
> Fetch the complete content index at: https://nordisk-ai.ghost.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# Most Companies Fail at Building Their Own AI Agents
- URL: https://nordisk-ai.ghost.io/most-companies-fail-at-building-their-own-ai-agents/
- Published: 2024-10-26T18:13:57.000Z
- Updated: 2026-08-27T17:21:08.000Z
- Description: Building your own AI agents can be a major challenge for many organizations, experts warn. Many companies fail because of the complexity involved and a lack of specialist expertise. Partnering with external AI vendors could be a more successful strategy.
- Author: Anders Hagen
- Tags: Application Areas, AI, AI Agents, automation, data architecture, digital transformation, Generative AI, machine learning, MLOps, open AI models, Sweden, Sweden, #Migrated-1787841634512, #wp, #wp-post, #Import 2026-08-27 14:41, #Import 2026-09-07 13:41

Agentic AI tools, designed to perform specific functions within an organization without human involvement, are becoming increasingly popular in the corporate world. Many organizations see these as a way to automate workflows, boost productivity, and tap into the potential of generative AI technologies. Despite this, experts warn against trying to build these agents in-house without the right expertise.

Analyst firm Forrester has named AI agents one of the top ten emerging technologies of the year, but it also issues a warning: Don't try to do it alone. According to Forrester's predictions for 2025, three-quarters of organizations that try to develop AI agents on their own will fail. Companies that don't succeed will instead turn to external AI consulting firms or use agents built into their existing software.

"Smarter companies will understand the current limitations and rely on their vendors and system integrators to build agents that are at the cutting edge of the technology," write analysts Jayesh Chaurasia and Sudha Maheshwari of Forrester.

## Complex Work That Requires Specialist Expertise

Building AI agents is a complex process that requires advanced AI expertise, something many organizations lack. The architectures behind these agents are complicated and involve multiple models, advanced RAG stacks (retrieval-augmented generation), sophisticated data architectures, and specialized knowledge.

Despite this, some companies do choose to develop their own AI agents. Goldcast, a software developer focused on video marketing, has experimented with a dozen open AI models to handle different tasks. The goal is to link these models together and turn them into agents that can carry out their tasks without human involvement.

"It's a more effective way to harness the power of agentic AI than building it from scratch," says Lauren Creedon, product manager at Goldcast. She adds that it takes advanced teams to pair different open models together into a working workflow, and that many companies will need to bring in outside specialists.

## The Benefits of Partnering With an AI Vendor

Senthil Kumar, CTO and head of AI at Slate Technologies, believes it's possible to develop your own AI agents, but it requires careful oversight and continuous improvement. "It's a collaborative process between the entire AI ecosystem and the human participants," he says. People need to be involved to review results and find ways to improve the agents' performance.

For many companies, it's a difficult tradeoff between building their own AI agents or partnering with a consultant, says Chris Ackerson, head of AI at AlphaSense. Large companies may be tempted to build their own agents, but they can easily stumble over fragmented data, underestimate resource needs, and lack the in-house expertise required.

"While some companies can achieve success, it's common for such projects to become very costly and complex," Ackerson says. Buying a solution from a trusted partner can help companies avoid pitfalls and speed up their path to success.

## Specialists Can Improve the Odds of Success

Adnan Masood, chief AI architect at UST, warns against trying to reinvent the wheel when it comes to complex systems like agentic AI architectures. Building these systems requires deep expertise in machine learning, natural language processing, and data engineering. By partnering with specialists or using ready-made solutions, companies can benefit from the experience that already exists in the field and thereby improve their odds of success.