BWA Agent Accelerator FrameworkPricing

One monthly fee. No per-seat licensing.

A fixed setup fee, then one all-in monthly that carries Azure consumption, the framework licence and support. Your bill tracks how much the assistant is used, not how many people you employ.

Per-user licence fees, at any tier

Monthly line: Azure, licence and support

Any time

Move the infrastructure to your own Azure subscription

Get A Quote Start With A Pilot

The Tiers

Priced on conversation volume, the model you run and whether your infrastructure is shared or dedicated. Monthly is all-inclusive: Azure consumption, framework licence and support. Figures in USD.

Tier	Conversations / mo	Setup	Monthly, all-in	What’s included
PoC / Pilot	up to 1,000	$8,000–12,000	from $300 / agent	Shared Azure Foundry, multiple models; CLU free tier, basic search, support
Starter	1,000–5,000	$15,000–25,000	$1,500–3,000	Shared Azure Foundry, multiple models, search and retrieval, support
Professional	5,000–25,000	$50,000–75,000	$5,000–10,000	Dedicated Azure Foundry, multiple models, CLU S tier, hybrid retrieval, priority support
Enterprise	25,000+	$100,000–150,000	$12,000–25,000	Dedicated Azure Foundry, multiple models, full search, SLA, dedicated support
Custom	negotiated	negotiated	negotiated	Complex requirements, multi-region, compliance

Dedicated infrastructure starts at Professional. Below that, your environment is isolated within a shared Azure footprint, which is what makes a pilot inexpensive enough to be worth running before you commit.

What the setup fee buys

Itemised, because a single implementation number tells you nothing about what you are getting. Most deployments need little custom development, the work is Azure configuration, loading your knowledge and training intent.

Item	Range	What it covers
Azure infrastructure setup	$2,000–5,000	Provision Azure resources, configure services
Branding and UI customisation	$2,500–7,500	Logo, CSS, front-end styling
Knowledge base setup	$3,000–15,000	Data load and document indexing. The widest line, and the one that decides answer quality — priced as the work it is
CLU training	$2,000–8,000	Intent and entity configuration, utterance training
Configuration and testing	$2,500–7,500	Azure config, integration testing, UAT
Total setup	$12,000–43,000	Typical deployment

Two shapes sit behind these numbers. Full is production grade, intent classification, entity extraction, hierarchical search, contextual imagery and stands up in one to two weeks. Lite is a proof of concept on Azure AI Search alone, retrieval without intent classification and no complex indexing: one to two days. Lite is the honest shape for a pilot.

One thing to plan for. Answer quality tracks the knowledge-base line above, not the tier you buy. Curating what the assistant is grounded in is the work that decides whether it is useful, so it is priced as work, not bundled out of sight.

What it costs over time

Year one carries the setup fee. Year two onward is monthly only.

Tier	Year 1	Year 2+ annual
PoC / Pilot	$11,600–24,000	$3,600–12,000
Starter	$33,000–61,000	$18,000–36,000
Professional	$110,000–195,000	$60,000–120,000
Enterprise	$244,000–450,000	$144,000–300,000

At 100,000 conversations a month

Three-year totals, and what each figure includes

Ours includes setup, framework licence, Azure infrastructure and support. The competitor figures are platform fees only, they exclude implementation, custom development and, where applicable, the cloud infrastructure and model usage billed separately. Read section 05 before the table.

Platform	Year 1	Year 2+	3-year total
BWA Agent Accelerator	$244,000–450,000	$144,000–300,000	$532,000–1,050,000
Copilot Studio (500 users)	$1,300,000	$1,200,000	$3,700,000
watsonx Assistant	$250,000	$150,000	$550,000
Dialogflow CX + Vertex	$150,000	$100,000	$350,000
Amazon Lex + Bedrock	$180,000	$120,000	$420,000

At 500 users the gap is structural, not efficiency. Copilot Studio’s three-year total of $3.7M against our $532K–1.05M is a per-user licence at $200 a month, multiplied by headcount. At 50 users the two are level; the difference is what happens as your organisation grows.

Three-year total at 100,000 conversations per month

Bars proportional to the largest total. Ours shows its floor solid and its ceiling pale.

BWA Agent Accelerator

$532K – $1.05M

Copilot Studio (500 users)

$3.7M

watsonx Assistant

$550K

Dialogflow CX

$350K

Amazon Lex

$420K

Solid = our floorPale = our ceilingSteel = platform fee only

Two of these are lower than our floor. They are platform fees with implementation, custom development and cloud infrastructure still to add, which is the comparison worth making, and the one section 05 sets out.

What each platform bills you for

The unit matters more than the rate, because it decides what makes your bill grow. Each vendor’s own published figures.

Platform	Billed on	Rate	Notes
BWA Agent Accelerator	the system	fixed setup, then one monthly	Azure consumption, framework licence and support in one line. No user count anywhere in it.
Microsoft Copilot Studio	per user	$200 / user / mo	Plus AI Builder capacity and Dataverse storage. One hundred users is $20,000 a month in licences.
IBM watsonx Assistant	monthly active users	free / $140 / custom	Lite covers 1,000 monthly active users; Plus is $140 per 1,000.
Google Dialogflow CX	per request	$0.007 / turn	$0.001 design-time, $0.007 runtime text, $0.001 per second of audio. Vertex AI is billed separately.
Amazon Lex	per request	$0.00075 / message	$0.004 for speech, $0.0065 streaming. Bedrock for the model is billed separately.

Per-user pricing grows with headcount, so you pay for people who never open it. Per-request pricing grows with every turn of every conversation, and bills the model separately. Ours grows with volume, in one line, and you can see the Azure inside it.

Scope, and how to read the comparison

The part most pricing pages leave out.

What the monthly covers

Azure consumption, the framework licence and support, one line, all-inclusive, with no per-seat component.

What is separate

Setup is a fixed one-time project fee, quoted from the table above. Engineering work beyond it, new integrations, custom development is hourly or on retainer.

How the comparison is built

Our figures include setup, licence, Azure infrastructure and support. The competitor figures are platform fees only: they exclude implementation, custom development and, where applicable, the cloud infrastructure and model usage those platforms bill separately.

On the comparison table

On platform fees alone, Google and AWS come in below our floor. We show that rather than omit it their line is a platform fee, ours is a delivered system, and the two are only comparable once implementation is priced in.

You can take the infrastructure with you

It is built entirely on Azure, so all of it can move to your own subscription at any time. None of the alternatives above can offer that their platform is the product, so leaving it means rebuilding.

Option	What it means	Best for
BWA managed	We manage the Azure subscription and billing	You want it turnkey
Customer managed	Azure resources transfer to your own subscription	You want full ownership
Hybrid	You own the infrastructure, we provide support	A balance of ownership and expertise

Why it costs what it costs

No per-seat licensing

Costs stay consumption-shaped and predictable. Adding people to your organisation does not add to the bill.

Less to build

The pre-built framework saves an estimated 60–80% of development effort and reaches production sooner.

Lower maintenance overhead

Standard Azure services throughout. No proprietary platform expertise required to keep it running.

Model flexibility

Swap AI models as better ones appear, with no migration cost for component upgrades.

Start with a pilot, not a signature.

A proof of concept stands up in one to two days on shared infrastructure. You will know whether the answers are good enough before anyone commits to a tier.

Get A Quote Start With A Pilot