Amazon Nova vs. IBM watsonx.ai

Where IBM watsonx.ai is well suited, based on our team experience: Num#1: Internal knowledge assistants for large companies. Like GAIA already described, is a good example. IBM watsonx.ai works very well as the base for a RAG solution. Num#2: Projects where data privacy is important, because we know that with IBM watsonx.ai client data are not used to train the models and everything stays inside IBM Cloud. Num#3: Solutions that need flexibility on models or projects designed to last a long time, because models change during time, so it's important to use a service that permit a smooth change of model Less appropriate: Num#1: Large and complex agent-based systems. Because IBM watsonx.ai gives the models and the APIs, but you have to build everything else yourself. So agent logic, orchestration, tool calls, etc. For GAIA this was manageable, and building our own multi-agent layer gave us full control, it was ok. But for larger projects, with many agents, complex workflows and integrations with many business systems, this means a lot of custom development and maintenance. So better to use other services (like orchestrate)

Incentivized

Invited by the Vendor

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. More Info

• It allows specialists to apply several base models for specific subtasks in the field of NLP.
• Gives the availability of many models developed for AI enhancement for different solutions.
• Has incorporated functionality for data governance and security to support access to AI tools by multiple users.

Incentivized

Invited by TrustRadius (TR) on Vendor's Behalf

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. More Info

• Should allow flexibility to handle unstructured data in a more robust data labeling and annotation tools
• real-time collaboration features are minimal which causes development less efficient.
• There is no clarity around automated model monitoring and retraining features
• Lack of pre-built industry specific models requires more customization effort for certain use cases

Incentivized

Invited by the Vendor

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. More Info

I still don't have enough experience, but i have seen a lot of demos and i have made some real world scenarios and so far so long every thing looks fine. I was at IBM Think 2025 and IBM TechXchange 2025 and the labs were really usefull and simple to understand.

Incentivized

Invited by TrustRadius (TR) on Vendor's Behalf

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. TR verified that a representative sample of customers was invited. More Info

I needed some time to understand the different parts of the web UI. It was slightly overwhelming in the beginning. However, after some time, it made sense, and I like the UI now. In terms of functionality, there are many useful features that make your life easy, like jumping to a section and giving me a deployment space to deploy my models easily.

Incentivized

Invited by the Vendor

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. More Info

I still don't have enough experience, but i have seen a lot of demos and i have made some real world scenarios and so far so long every thing looks fine. I was at IBM Think 2025 and IBM TechXchange 2025 and the labs were really usefull and simple to understand.

Incentivized

Invited by TrustRadius (TR) on Vendor's Behalf

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. TR verified that a representative sample of customers was invited. More Info

I think IBM watsonx.ai is more IBM products focused so it helps find information for IBM related products better. For example I would with IBM assembler, but this language doesn't have too much documentation in GitHub Copilot, so the results I get from it aren't as helpful

Incentivized

Invited by TrustRadius (TR) on Vendor's Behalf

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. More Info

I still don't have enough experience, but i have seen a lot of demos and i have made some real world scenarios and so far so long every thing looks fine. I was at IBM Think 2025 and IBM TechXchange 2025 and the labs were really usefull and simple to understand.

Incentivized

Invited by TrustRadius (TR) on Vendor's Behalf

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. TR verified that a representative sample of customers was invited. More Info

• For one of the usecase we were able to save 1.3 million dollar by creating the COB project in 2 months by leveraging the whole ecosystem and savibg of approximately 2800 hrs per annum
• For the icd usecase wherre we have used the graph memgraph using docker to deploy it and created graphs and then created inference pipeline saved 20 minutes per medical coder and approximately of 3.5 million dollar savings as claims submission verifying on the payer side is a lengthy process

Incentivized

Invited by TrustRadius (TR) on Vendor's Behalf

A de minimis incentive was given to thank the reviewer for their time. The incentive was not used to bias or drive a particular response, nor was the incentive contingent on a positive endorsement. TR verified that a representative sample of customers was invited. More Info