Consolidated vs. best-of-breed AI strategies: A C-suite guide

Consolidated vs. best-of-breed AI: Key differences

The way a business acquires its AI technologies has major implications across the following areas:

Procurement complexity

For starters, procuring AI tools is easier if the business pursues a consolidated strategy. In that case, it only needs to evaluate the overall platforms of each vendor it's considering working with.

A best-of-breed strategy is more complex from a procurement standpoint because it requires assessing how competing AI offerings perform across a variety of use cases. It might be necessary to reevaluate vendors frequently, given the rapid pace at which AI tooling options -- and the business needs they address -- evolve.

Total cost of ownership (TCO)

Neither a consolidated nor a best-of-breed AI tool architecture is necessarily better on the total cost of ownership front, but each offers different opportunities to optimize costs.

With best-of-breed strategies, businesses have greater flexibility in finding the most cost-effective AI tools for specific use cases. This can result in lower TCO, provided the business is effective at comparing the costs of competing offerings and determining which will address business needs at the lowest cost.

The cost advantage of a consolidated AI platform strategy is that vendors might offer discounts to enterprise customers who commit to using their platforms at scale. The exact terms of enterprise AI pricing discount agreements vary, however, and business leaders shouldn't assume that a consolidated platform approach will yield steep discounts or prove more cost-effective than a curated selection of best-of-breed tools.

No matter which type of AI strategy a business adopts, the same operational best practices apply for keeping AI costs in check. Those practices are likely to play a greater role in determining AI TCO than the business's chosen AI strategy.

Performance

Best-of-breed AI stacks tend to perform better than consolidated AI platforms. But whether a business will see better performance depends on how effectively it selects its bespoke AI tools under a best-of-breed strategy -- and how effectively it updates its AI stack as newer, better options emerge.

Implementation and maintenance

From an implementation and maintenance perspective, a consolidated AI strategy wins in most respects.

A consolidated platform has fewer integrations and variables, making it easier to set up. AI technology vendors can also provide forward-deployed engineer (FDE) teams -- which specialize in implementing and, in some cases, maintaining enterprise AI -- to support clients who purchase AI products at scale or exclusively from them.

Getting FDE support for AI deployments focused on a single use case or business domain is not impossible, but it's harder due to the smaller purchase volume.

Scalability

Scaling AI is easier with a consolidated AI stack. That's because the business can expand the scope of its AI use without first procuring and implementing new tools.

That said, best-of-breed tooling can be easy to scale as well, so long as scaling doesn't require introducing new tool types. For example, if the goal is to scale a pilot project to a production deployment, businesses can do that without new technology because the platform that powered the pilot can also power the production deployment. But addressing a new type of use case might require adding an AI tool or platform tailored to that novel need, which takes time.

Interoperability

From an interoperability perspective, best-of-breed stacks are more flexible overall. When businesses create AI architectures to span multiple platforms and vendors, it's simpler to add new ones without overhauling the underlying AI stack.

However, the modern AI ecosystem has evolved to the point where AI vendors are making their offerings accessible to third parties. As a result, a consolidated AI platform strategy could also offer a fair amount of interoperability. An enterprise could connect a new type of AI agent to its platform, regardless of the vendor or model it uses, because most agents are compatible with mainstream platforms.

Governance capabilities

Governance can be simpler, but less granular and less effective, under a consolidated AI strategy. A consolidated strategy lets businesses lean on a single vendor's set of governance controls, which is easier than working with governance tooling across multiple platforms. However, it might also reduce governance flexibility because the business must settle for whichever governance controls its vendor supports. 

Security

Security, too, is simpler for businesses with a consolidated AI strategy. Greater consistency translates to fewer vulnerabilities that attackers could exploit. This doesn't mean, however, that it's impossible to secure a best-of-breed AI stack. It's feasible, provided businesses deploy appropriate security controls across all the AI platforms they use.

Like cost management, security management is an aspect of enterprise AI in which outcomes are shaped more by how well businesses secure their AI infrastructure and assets than by the type of AI architecture they choose.

The role of model portability

Beyond weighing the pros and cons of consolidated vs. best-of-breed AI strategies, an additional critical consideration for businesses deciding which approach to take is the extent to which they can benefit from model portability.

Model portability is the ability to move AI services, applications and agents from one model or model provider to another. If the same AI agent can run using either OpenAI or Anthropic models, it's a model-portable agent.

Model portability, from the perspective of choosing an AI strategy, makes it easier to implement a best-of-breed approach. When any model or vendor can power a business's AI stack, it becomes feasible to build out a stack of discrete AI services and tools, and to update them without worrying about breaking workflows.

For businesses that opt for a consolidated AI strategy, model portability helps protect against changes that their chosen vendor might make to its models. Model deprecation is less likely to cause problems if the business's applications can switch to use a different model.

How to enable model portability

Achieving model portability depends on the types of AI tools a business builds or buys. When developing AI applications or services in-house, practices like creating model-agnostic interfaces, choosing standardized APIs or using abstraction layers to decouple applications from models help ensure portability.

As for third-party offerings, many mainstream AI applications and agents are model-portable by default, or at least compatible with models from all major AI vendors. OpenAI and Anthropic can power most of Salesforce's agentic AI capabilities. Likewise, although Microsoft Copilot uses OpenAI by default, users can configure it to work with Anthropic instead.

AI tools that address more bespoke needs, or tools that an enterprise develops in-house, might work only with a specific model or model family, leading to a lack of model portability. A product might also work with all major models but not with lesser-used models or vendors. Procurement teams should pay attention to these details to ensure that they select options that deliver the right level of model portability.

What to choose when: A checklist for AI procurement decisions

Businesses must determine the best AI strategy based on both their near-term operational objectives and long-term strategic goals. To decide which AI strategy -- consolidated and/or best-of-breed -- makes most sense, consider the following key factors:

• Current and future AI use cases. If a consolidated AI platform is not broad and flexible enough to support all the business's intended AI use cases, a best-of-breed strategy might be best.
• State of AI maturity and growth plans. Businesses whose AI strategies are already mature -- in the sense that they've brought AI into production and have a concrete sense of what they need -- can more easily adopt a consolidated platform. For those still experimenting, a best-of-breed architecture offers more flexibility.
• Internal engineering resources. Businesses with limited in-house AI or data science expertise might be better suited to a consolidated AI platform, which is easier to implement and manage.
• Regulatory and governance requirements. The relative simplicity of a consolidated platform is a boon to businesses facing complex compliance or governance mandates. That said, a best-of-breed strategy can offer greater flexibility by, for example, enabling the use of different models across different geographical regions in response to each jurisdiction's distinct compliance requirements.
• Budget and discount goals. For businesses with tight budgets, a consolidated platform might yield greater cost savings, provided they can negotiate discounted rates. Businesses with smaller levels of AI consumption might not be large enough customers to secure discounts, even if they commit to a single vendor.
• Tolerance for vendor concentration. A consolidated AI strategy requires putting all the business's eggs in one AI basket. This might be fine for businesses that trust their chosen AI vendor or have high model portability and can switch to a new vendor if necessary. Those worried about lock-in, however, might be better served by a best-of-breed strategy.

Chris Tozzi is a freelance writer, research adviser, and professor of IT and society who has previously worked as a journalist and Linux systems administrator.