Amazon Bedrock · Small companies 3 weeks · $10,000 fixed · Your AWS account

Amazon Bedrock for small companies, without an ML team.

You're a small company on AWS, and you want to know whether Amazon Bedrock can do a real job in your business. The Bedrock Feasibility Sprint answers that in three weeks: a working proof of concept in your own AWS account, a cost model, and a written go/no-go.

Price
$10,000, fixed
Duration
3 weeks
Your time
2–3 hours / week
AWS costs
Typically under $200
Why small companies

You can't spend a quarter finding out if it works.

Large companies have AI teams and time to run long evaluations. A small company usually has one technical lead and a budget that has to show a return.

The Sprint is sized for that. It tests one use case on your own data, at a fixed price. You don't hire anyone or sign a long contract, and you finish with evidence you can act on.


Example use cases

Problems a three-week Bedrock proof of concept can test.

The first two are patterns from my own client work on Amazon Bedrock. The Sprint tests one use case at a time.

Answers across separate systems

Plain-language questions about project status, costs, or timelines, answered from systems that don't talk to each other today.

Routing questions to the right data

An agent that takes a question from staff or customers and sends it to the right source: orders, customers, assets, or schedules.

Search and answers over your documents

Policies, contracts, manuals, or past proposals, searchable in plain language with answers that cite the source document.

Support ticket triage and drafts

Incoming tickets or emails sorted and summarized, with a draft reply for a person to review before it goes out.

What you leave with

Evidence you can take to the owner or the board.

  • A working proof of concept on Amazon Bedrock, in your AWS account, deployed with CloudFormation.
  • A cost model with actual proof-of-concept costs and a projection at production volume.
  • A 2–4 page go, adjust, or no-go memo against criteria we agree in week one.
  • A risk register covering data quality, model choice, evaluation, and compliance.

See the full Sprint, week by week

Is this for you?

Who the Sprint suits.

If it's not a fit, I'll say so on the scoping call and point you somewhere better.

Good fit if you…

  • Are a small company or startup already running on AWS.
  • Have one specific use case in mind.
  • Have sample data, even if it's messy.
  • Can give a technical lead 2–3 hours a week.
  • Want an answer before you spend on a full build.

Not a fit if you…

  • Want to compare Bedrock against Azure OpenAI or Vertex AI.
  • Need a finished chatbot interface at the end of three weeks.
  • Don't have any sample data to test with.
  • Haven't picked a use case yet.
Questions small companies ask

Before the scoping call.

How much does Amazon Bedrock consulting cost for a small company? +

The Bedrock Feasibility Sprint is $10,000, fixed, for three weeks. AWS infrastructure for the proof of concept is billed to your own account and is typically under $200 for the Sprint window.

Do we need a machine learning or data science team? +

No. You need a technical lead who can give 2 to 3 hours a week for working sessions, a specific use case, and sample data. I do the build.

What do we get at the end? +

A working proof of concept in your AWS account, deployed with CloudFormation, a cost model at proof-of-concept and production scale, a written go, adjust or no-go memo, and a risk register.

Who owns the code? +

You do. Everything ships into your AWS account and your repository. There is no consultant infrastructure to migrate off later.

What happens after the Sprint? +

If the answer is go, a production proof of concept runs $25,000 to $50,000 and is scoped separately. If the answer is no, you have a written reason and you've spent three weeks instead of months. For ongoing AWS help, the OrbitOPS monthly retainer covers 40 hours a month.

Find out if Bedrock fits

Send a paragraph about the problem.

Tell me what you're trying to do, what data you have, and what a good result would look like. I'll come back with a Sprint scope, a different recommendation, or a pointer to someone better suited.