--- title: "How agencies deliver AI systems to clients" description: "Plan an agency delivery process for client AI systems, from task selection and source access to acceptance checks, team training, and a clear handoff." canonical: "https://scalewithsearch.com/articles/how-agencies-deliver-ai-systems-to-clients" --- ## Site navigation - [Scale With Search](https://scalewithsearch.com/) - Services - Services - [AI systems](https://scalewithsearch.com/ai-systems) - [Site build](https://scalewithsearch.com/work#site-build) - [Whole website fix](https://scalewithsearch.com/work#whole-website-fix) - [Content library with SEO](https://scalewithsearch.com/work#content-library) - [Working session](https://scalewithsearch.com/audit) - Start here - [Send your brief](https://scalewithsearch.com/work#send-your-brief) - [Example scope](https://scalewithsearch.com/work#review-the-example-scope) - [Prices and terms](https://scalewithsearch.com/work#prices-and-terms) - Industries - Real estate - [Real estate agents and teams](https://scalewithsearch.com/for/real-estate) - [Platforms, ownership, and IDX](https://scalewithsearch.com/for/real-estate/real-estate-website-platforms-ownership-idx) - [Search, content, and AI](https://scalewithsearch.com/for/real-estate/real-estate-search-content-ai) - [Inquiries, campaigns, and handoffs](https://scalewithsearch.com/for/real-estate/real-estate-website-inquiries-campaigns-handoffs) - [Buyer, seller, and local pages](https://scalewithsearch.com/for/real-estate/real-estate-buyer-seller-local-resources) - [Rules, privacy, and trust](https://scalewithsearch.com/for/real-estate/real-estate-website-rules-privacy-trust) - Browse by business - [Trades and home services](https://scalewithsearch.com/for/#group-trades) - [Health, wellness, and animal care](https://scalewithsearch.com/for/#group-practices) - [Law, finance, and IT](https://scalewithsearch.com/for/#group-professional) - [Marketing and SEO agencies](https://scalewithsearch.com/for/#group-agencies) - [Shops and manufacturers](https://scalewithsearch.com/for/#group-manufacturing) - [Restaurants, shops, studios, and nonprofits](https://scalewithsearch.com/for/#group-local-services) - Chiropractic practices - [Chiropractic and wellness practices](https://scalewithsearch.com/for/chiropractic-and-wellness-practices) - [Chiropractic and physical therapy clinics](https://scalewithsearch.com/for/chiropractic-and-physical-therapy-clinics) - [All industries](https://scalewithsearch.com/for/) - Resources - Free tools - [AI readiness check](https://scalewithsearch.com/check) - [Free tools](https://scalewithsearch.com/tools/) - [Free Starter Kit](https://scalewithsearch.com/kit/business-memory-starter-kit.zip) - Guides - [AI-native library](https://scalewithsearch.com/ai-native) - [Industry intelligence](https://scalewithsearch.com/intelligence) - [Office job guides](https://scalewithsearch.com/guides/) - [Business memory](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [Agent governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [Migration and ownership](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - Articles - [All articles](https://scalewithsearch.com/articles/) - [What changed](https://scalewithsearch.com/changes/) - About - Evidence - [Proof](https://scalewithsearch.com/proof) - Scale With Search - [About](https://scalewithsearch.com/about) - [How it works](https://scalewithsearch.com/how-it-works) # How agencies deliver AI systems to clients An agency can deliver an AI system by taking responsibility for one defined business function. Start with the client's work, agree on the result, build within approved access, and teach the client team to run it. Make acceptance depend on observed behavior. This article proposes a delivery process. The examples describe possible scopes, not completed client projects. The focus is the agency's relationship with its client throughout the build. For the underlying account boundary, use the separate guide to [keeping client context apart](/articles/separate-client-context-before-agent-work). ## Start with a function the client can name Ask the person doing the work to show you the last completed example. Follow the source material through preparation, review, correction, and delivery. Record where someone makes a decision and where someone copies or rearranges information. Suppose the example is a weekly account update. The inputs might be an approved project record and a list of open questions. The output might be a draft update for the account manager. Sending the update is a separate action with its own approval. Keep the proposed scope narrow enough to describe in one sentence: prepare the weekly update from these records for this reviewer. Avoid promising a business transformation before you understand that path. A small scope still needs a complete start, middle, and finish. Ask what would make the update unusable. A wrong client name, unsupported delivery date, or missing blocker should become a test case. The client's rejection criteria belong beside the desired output. ## Sort the surrounding work before building List the recurring steps around that function. Mark each eliminate, automate, augment, or outsource. Use the list to decide which work should remain in the process. A duplicate status document may be a candidate for elimination. Copying a validated project identifier may suit a fixed rule. Drafting a short explanation may suit AI with review. A specialist decision may belong with an outside professional. Treat those classifications as proposals for the client to check. Ask the team why a seemingly redundant step exists. It may satisfy an agreement or support another department. Remove it only after its purpose and downstream readers are understood. The [AI systems method](/ai-systems) follows this study and sort before building one function. The agency can use that sequence inside its own business first. Keep any demonstration free of another client's records. ## Agree on a delivery brief Write a brief that names the trigger, inputs, output, reviewer, and stopping point. Include the records the system may read and the accounts it may use. State what the client must supply before testing begins. For the weekly update example, the client might provide a current project list, approved wording rules, a sample accepted update, and a reviewer. The agency supplies the function, operating instructions, tests, and training named in the scope. Define the handoff conditions before choosing tools. The [build acceptance list](/articles/what-a-build-acceptance-list-contains) provides a fuller structure for connecting requirements to evidence. Use it to turn vague phrases such as ready to use into observable checks. NIST describes its [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) as voluntary guidance for AI design, development, use, and evaluation. It supports examining risk across delivery. It does not certify an agency's implementation. Source checked October 26, 2026. ## Keep the client relationship clear Name one agency owner for delivery and one client owner for acceptance. Add a technical owner when another firm controls the systems involved. Record which person can approve data access, scope changes, and production use. If the agency brings in a builder, explain the arrangement before sharing records. Specify who sees the client's material, who maintains the function, and who receives support requests. Do not leave the client to discover another provider through an account invitation. Keep responsibility for communication with the named relationship owner. The builder should be able to report a technical blocker without making a new business commitment. A client request that changes the result returns to the agency owner for a written scope decision. Use the [agency context structure](/articles/agency-brain-retains-client-context) to separate reusable methods from each client record. Put those decisions in a shared project record. Private messages can inform the work, but the accepted instruction should be readable by the people responsible for delivery. ## Build the preparation path first Create a version that uses approved sample records and produces a draft. Keep external sending, publishing, and account changes outside that first path. The team should be able to inspect the inputs and compare the draft with the accepted example. Use deterministic checks before asking AI to interpret material. Resolve the client identifier, verify required inputs, and reject sources outside the allowed set. Let the model handle the part that requires language or judgment within those boundaries. For example, a fixed check can reject an update request without a project identifier. AI can then organize the approved project's notes into a draft. The reviewer decides whether the explanation matches the source record and is suitable for the client. Record unresolved questions in the output. A missing delivery date should remain missing. It should not become an estimate because the draft reads better with a date. ## Make review part of the work Use 10/80/10 as a design shorthand: a person sets direction, the system prepares the middle, and a person checks the result. The percentages describe the method. They are not a measured allocation of effort or a time-saving promise. Give the reviewer the source references beside the draft. Ask them to check facts, omissions, recipient, and the next proposed action. Capture a correction in the client-owned record so the next run can read the accepted rule. Separate approval of the draft from permission to send it. [OWASP's excessive-agency guidance](https://genai.owasp.org/llmrisk/llm062025-excessive-agency/) recommends limiting capabilities and enforcing authorization in connected systems. Apply that principle at the tool that performs the action. Source checked October 26, 2026. For this example, acceptance could require that a draft-only account cannot send an update at all. A written instruction alone should not carry that boundary. ## Show cost and failure behavior Name a monthly AI spending limit for the function and a person who receives the alert. Write down what happens at the limit. A draft should not silently trigger repeated retries until someone checks a bill. Require a receipt for every run. Include the client identifier, input references, output location, model identifier, cost, and final state. Use completed, needs review, blocked, or failed consistently. A missing source should produce a visible blocked result. Keep model selection in documented configuration. Test any replacement against the same accepted examples before changing production. Swapping a model should preserve the agreed result and limits, although the wording may change. Distinguish the agency's delivery fee from ongoing model, hosting, and subscription charges. The client should know which accounts carry those costs and who may change the limits. ## Rehearse the handoff with the client Ask a client team member to run the function from its written instructions. Have them locate the source, start a sample run, review the draft, apply a correction, and find the receipt. Observe where they need undocumented help. Add the missing instruction and repeat the affected step. Include a failed-input exercise and the stop procedure. A polished walkthrough is useful preparation, but acceptance should depend on the client's attempt. Close the delivery record with accepted checks, open exceptions, account ownership, and the support boundary. State who handles a broken connector and who handles a new requested function. A future client should receive a fresh scope and separate records, even when the agency reuses its delivery method. ## Send one client function for scope Choose a recurring job your agency can explain from input to review. Bring its current example, source locations, and acceptance owner. Scale With Search scopes [AI systems](/ai-systems) by working function, including the owned record and team handoff. [Send your brief](/work#send-your-brief). ## Related: Ownership and Migration - [Separate Client Context Before an Agent Touches the Work](/articles/separate-client-context-before-agent-work) - [What a Build Acceptance List Contains](/articles/what-a-build-acceptance-list-contains) - [How to Build an Agency Brain That Retains Client Context Without Mixing Accounts](/articles/agency-brain-retains-client-context) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```