# CX Foundry # You don't need a platform.You need a process.Then decide. Your customers want a better experience. Your board wants AI in the plan. The market wants you to pick a platform. **That’s the trap.** [Start with a Snapshot](/get-started) [How we work](/how-we-work) [Try the demo](/demos/cxfoundry) --- 01**The problem ## Three reasons GenAI service projects fail ### Wrong tech. Your CCaaS vendor's GenAI module looked good in the demo. But it can’t handle the edge cases in your actual conversations. You find out in week sixteen. ### Shallow discovery. You build something that works. But it falls apart against the live mix of customer needs, messy data, and policy constraints. You find out in QA. If you’re lucky. ### Compliance kills it. Regulatory, policy, and infosec issues aren’t considered early enough. Risk and compliance pick them up two weeks before go-live. The project stalls. These aren’t platform failures. They’re process gaps that used to be hard to fix. **CX Foundry changes that.** --- 02**The AI loop ## The AI loop CX Foundry is an automated AI delivery loop that gets you from call recordings, to proven AI agents capable of serving your customers. **Accountable at every step****Your Principal and your team guide the work ConversationsEvidenceAI agentSimulationEvaluationStage gate Conversations Collect a representative sample of contact center recordings, chats, or transcripts. Go broad, and with enough volume to see repeated patterns. Evidence Conversation data analysis identifies opportunities. AI agent AI + your inputs + Principal expertise creates a working agent. Simulation Working agents are tested against simulated scenarios derived from conversation data. Evaluation Evaluation drives changes to the AI agent assets: prompts, tools, knowledge, and scope. Stage gate Fast decisions when the evidence says a workflow should be narrowed, stopped, or taken forward. We build AI, with AI. With you. It's a [reusable method](/method) that capable internal teams can use to refine their AI agent or tackle another use case with the work we do together and the AI tools that are now everywhere. --- 03**The path ## The path We apply the method in three stages, building stronger agents, and deeper proof at each stage. SnapshotDiscover 1–2 weeks Find the real opportunity - Opportunity sizing - First workflow prototype - 90-day portal access Stage 1Build 6 weeks Prove the agent works - Working AI agent - Workflow + tools - Value model Stage 2Prove 8 weeks Secure the launch path - Production readiness - Runtime package - Evaluation harness Deploy & RunOperate Ongoing Integrate and deploy. - Run in your cloud - Deploy on your platform - Prove with the eval harness Option one ### Run it yourself. Take the proven OpenAI agents, backend/tool layer, runtime components, and evaluation harness into your own AWS, Azure, or GCP environment. OpenAIAzure Option two ### Translate to a SaaS platform. Use the CX Foundry outputs to configure the same workflow on Parloa, Gemini Enterprise for CX, your existing CCaaS, or another specialist vendor. Gemini In either scenario, you’ll need to integrate the solution into your telephony, contact center, and backend systems and re-run the evaluation harness to make sure what goes into production delivers the outcomes CX Foundry proved from your data. [Start with a Snapshot](/get-started)[How we work](/how-we-work) --- 04**Who it suits ## Does this sound like you? [Start with a Snapshot](/get-started)[How we look after your data](/data) - You handle more than a million calls per year. - You have a named business sponsor with budget authority. - You want a credible path to production, not another isolated AI pilot. - You can share a sample of call recordings or transcripts. If you want a generic demo, can’t share a sample, don’t have a sponsor, or you’re just shopping for a platform swap — we’re not the right call, and we’ll tell you on the first one. --- Your data ## How we handle your data. CX Foundry starts with data. We need call recordings or transcripts to drive the process, but we understand that handling sensitive customer data is a big responsibility. That’s why we use a dedicated secure environment, process, and toolset to receive, transcribe, and redact your data — operated under Waterfield Tech’s ISO/IEC 27001-certified Information Security Management System. Read on to understand what we need from you and how we keep you, your customers and your data safe. Feel free to share with your information security team. Secure redaction environment** STEP 1 Share Credentials to an Azure Blob Storage location we provision for you. STEP 2 Transcribe Azure AI Services, inside the secure environment. STEP 3 Redact Names, IDs, card numbers, addresses, emails replaced with labels. STEP 4 Review A human checks for anything left. If found, redaction runs again. STEP 5 Analyze Approved AI tools only: Microsoft, OpenAI, Google, Anthropic. STEP 6 Delete 180 days after the engagement, from storage and backups. ### What we ask for **1,000+ recordings or transcripts** plus operating context. You keep all rights to your data. A representative sample of your customer interactions — typically 1,000+ audio recordings or text transcripts — plus enough operating context to make sense of them. You retain all right, title, and interest in your data. ### How you share it **Your own Azure Blob container.** Stereo audio or timestamped transcripts preferred. We securely share credentials to a Microsoft Azure Blob Storage location we provision for you. We’re flexible on format, but prefer high-quality stereo audio with separate channels for customer and service agent, or transcribed data with timestamps. ### Where it lives **Two client-isolated containers,** least-privilege access, every action logged. We provision two client-isolated Azure Blob Storage containers: one for non-redacted data, one for redacted. We apply least-privilege, role-based access. Only a small number of named delivery staff have access, and their actions are logged and monitored, with alerting on anomalies. Only data that has completed redaction and review is approved to leave the redaction environment. After that, it continues to be tracked and managed under our broader ISO 27001 information security management system, where it is analyzed by approved AI tools and infrastructure from Microsoft, OpenAI, Google, and Anthropic that Waterfield has vetted. ### Restricted data **Never sent to AI without written authorization:** secrets, card data, health data, or anything you flag. We will not input or process the following using any AI tool, including approved AI, without your explicit written authorization: passwords, API keys, private keys, tokens, payment-card data, sensitive personal data including health data, or any information you have labeled "SENSITIVE INFORMATION, MUST NOT BE PROCESSED BY AI" or equivalent. ### How long we keep it **180 days** after the engagement by default. Earlier on request, confirmed within 14 days. 180 days from the end of the engagement by default. After that, we securely delete from primary storage and from backups consistent with our backup rotation. You can ask us to delete earlier in writing; we will confirm completion within 14 calendar days. ### What we never do **No training on your data.** No sharing beyond approved providers. No use for any other client. We do not use or share your data to train, fine-tune, retrain, or improve any shared or general-purpose AI model. We only disclose it to approved providers where needed to deliver the engagement. We do not apply your data to work for any other client. We do not use your data outside the engagement you signed up for. We may retain and use generic learnings and anonymized insights derived from your data, only where it would be impossible to extract or reconstruct your confidential information from them. Operated under Waterfield Tech's ISO/IEC 27001-certified Information Security Management System --- 05**Example Global e-commerce and fintech platform ## The hard case: payment reminders. The hard case: reminding customers about missed payments on a retailer-issued credit line. Emotionally charged. Brand-sensitive. Traditionally handled by humans with inconsistent results. The CX Foundry method coupled with Parloa’s enterprise GenAI voice platform outperformed human agents. [Read the joint case study](https://www.parloa.com/customers/ai-powered-customer-care-transforming-payment-reminders-for-leading-e-commerce-platform/)[Waterfield on the Parloa partner directory](https://www.parloa.com/find-a-partner/partner/waterfield-tech) 51% 66% HumanAI agent Payment commitment +15 pts 57% 62% HumanAI agent Payment follow-through +5 pts Payment-reminder agent · human vs AI · percent of customers --- The engagement ## How we work CX Foundry is Principal-led, evidence-first, time-boxed, and fixed-fee. It is designed to get you from real customer conversations to proven OpenAI agents you can take live or translate to another platform. The full [CX Foundry method](/method) is public. The commercial engagement is how you get it delivered by a CX Foundry Principal with their judgment, expertise, and AI tools that ensure you get the outcome you want, with the accountability you need. Fast. ### The path Snapshot: find the real opportunity. One to two weeks. We analyze a representative sample of your calls, chats, or transcripts, identify repeated workflows, size the opportunity, and build a working voice AI prototype for the strongest candidate. Stage 1: prove the agent works. Six weeks. We turn two to three priority workflows into working agent assets, backend tools, test data, simulation runs, and an evidence-based value model. Stage 2: secure the launch path. Eight weeks. We take the strongest one or two workflows through broader simulation, adversarial testing, policy review, and risk analysis. We deliver a reusable evaluation harness and a small runtime package that can be used to host your agent using OpenAI voice or chat models on AWS, Azure, or GCP. Run it yourself or translate to a SaaS platform: Take the proven agent, backend tool layer, runtime components, and evaluation harness into your own environment with your team or partner of choice, or use the CX Foundry outputs to configure the same workflow on Parloa, Gemini Enterprise for CX, your existing CCaaS, or another specialist vendor. CX Foundry itself does not include last-mile telephony routing, production backend integration, live operations, or managed service support. It gives you the evidence, assets, and test harness to deploy and prove those things with confidence, on your own, with another partner, or with Waterfield Tech. ### Your role The work moves quickly because we keep the team small and the inputs clear. Success demands: - A named business sponsor with budget authority - A representative sample of customer conversations - Access to operations, technical, data, security, and compliance stakeholders at the points where their judgment matters - SOPs, policies, scripts, escalation rules, API documents, schemas, sample payloads, and test data - Fast decisions when the evidence says a workflow should be narrowed, stopped, or taken forward --- Commercial terms ## Payment model Our invoicing and payment model is different from a traditional project because the delivery model is different. We reserve a named Principal for a defined slot, use the CX Foundry method and tooling to compress the work, and keep the commercial backstop simple. With CX Foundry 50%to reserve Deposit reserves your Principal's delivery slot Delivery slots can be reserved with a 50% deposit. The remaining 50% is invoiced on stage acceptance. If you don’t accept the stage after a structured review process, the balance is not invoiced; ownership and use rights do not transfer, and the engagement closes for that stage. ### Why it works The model removes the coordination overhead of endless scheduling and rescheduling, scoping and re-scoping. You and your Principal share the same incentive: turn the available evidence into useful work and clear decisions. **50%**Deposit to reserve a delivery slot **50%**Invoiced on stage acceptance **$0**Balance if the stage isn't accepted --- Our team ## The Principals Your CX Foundry Principal is the human thread of accountability that gets you from calls and transcripts to proven AI agents. They have deep experience in customer service, AI, business transformation, and facilitation, which they leverage to deliver the CX Foundry method, using the AI-native tools we’ve built to empower them. Your Principal brings the judgment; CX Foundry's tooling gives them the reach and speed of a much larger team. With CX Foundry you pay for the output, not the coordination overhead. CX Foundry Principal ### Kerry Robinson Kerry works where customer experience, applied AI, and operating reality meet. He is strongest at turning ambitious AI ideas into inspectable systems: the evidence, agent instructions, workflows, and proof needed for a production decision. He cares about building the smallest thing that proves the right thing. [LinkedIn](https://www.linkedin.com/in/k3rryrobinson/)[Writing](https://thedualist.substack.com) CX Foundry Principal ### Sarah Reitsma Sarah helps enterprise teams make conversational and generative AI useful in the places where customers actually feel the system. Her work sits close to conversation design, healthcare CX, operating friction, and human trust. She is especially good at making complex service paths legible enough to improve. [LinkedIn](https://www.linkedin.com/in/sarahreitsma/) CX Foundry Principal ### Damian Mathews Builds website chatbots from company content and customer journeys. Brings a marketing perspective to positioning, conversion, and customer value. [LinkedIn](https://www.linkedin.com/in/damian-mathews/)[Writing](https://waterfieldtech.com/insights/?_resource_filters=the-last-mile) --- Public guide ## The CX Foundry method. CX Foundry is a method to transform raw contact center recordings or chat transcripts into proven AI agents that can reliably automate a portion of those calls or chats. We’ve empowered CX Foundry Principals to deliver the method, and built tools that allow them to delegate a lot of the detailed work to AI, while maintaining a golden thread of accountability from your data, documents, inputs, and decisions through to a safe, successful, live deployment. If you want to do the whole thing yourself, this page details every major step in the method. You’re welcome to take it and use it yourself, or paste the markdown version into your favorite AI and let it guide you through the steps. If you don’t already have the tech and talent to deliver this method reliably, we hope you’ll consider doing a CX Foundry Snapshot, Stage 1 and Stage 2 with us to get live faster and manage risk. HumanMarkdownLicense ### The simulation and evaluation loop is the heart of it - Conversation data analysis identifies opportunities. - AI + your inputs + Principal expertise creates a working agent. - Working agents are tested against simulated scenarios derived from conversation data. - Evaluation drives changes to the AI agent assets: prompts, tools, knowledge, and scope. - Then simulate and evaluate again. And again. Building from performance, to policy adherence, to infosec and compliance testing in a fast, iterative, AI-native loop. Simulation and evaluation starts in Stage 1 as soon as there is something worth testing. It becomes the center of Stage 2. ### Why it works - It starts from real conversations. - Each stage is justified by working assets from the previous stage. - The evaluation process optimizes for performance and compliance. - It defers a platform decision until the evidence is strong enough to justify it. SnapshotFind the opportunity11 steps - Collect a representative sample of contact center recordings, chats, or transcripts. Go broad, and with enough volume to see repeated patterns. - Transcribe the conversations and redact private information. The goal is to preserve what happened in the interaction while removing names, account numbers, health details, payment data, or anything the analysis process doesn’t need. - Normalize the data into a clean evidence set. Put each conversation into a consistent format with speaker turns, timestamps where available, metadata, call duration, channel, and any known outcome. - Analyze every conversation with the approved analysis model. Categorize the conversation, score friction or effort, identify the customer intent, capture the resolution path, and note where the process broke down or required human work. - Build clusters of similar conversations. Look for repeated patterns such as “status request,” “appointment reschedule,” “refund follow-up,” or “missing authorization.” - Size the opportunity. For each major cluster, estimate volume, average handle time, and automation value. Choose the cluster that is large enough to matter, repeated enough to automate, and simple enough to prove. - Infer the process behind the calls. Map what the human agent appears to do: what information they collect, what systems they check, what decisions they make, what policy rules they follow, and where they escalate. - Identify the tools the AI agent would need. This usually becomes a short list of backend actions: look up an order, check account status, search availability, verify eligibility, create a case, send a confirmation, or hand off to a person. - Build a realistic demo agent and backend stub. Use AI code generation to create a working agent and a realistic fake API so stakeholders can experience the future workflow before production systems are touched. - Build a simulated customer agent to test the demo. Have one AI agent play realistic customers against the demo agent, including normal cases and edge cases, then iterate until the agent completes the workflow or hands off safely. - Demonstrate the experience to stakeholders. Show the evidence, the opportunity size, the live agent interaction, and the proof from simulated runs so business, operations, and technical teams can align on whether the workflow should move to Stage 1. Complete whenThe snapshot is complete when the team can say: “Here is a repeated customer conversation pattern, here is why it matters, here is the process behind it, here is the agent we can build, and here is the evidence that it is worth proving properly.” In rare cases, a snapshot might end with: “No such opportunity could be found”, which is just as valuable. Stage 1Prove the agent works8 steps - If necessary, gather additional conversation data and repeat the analysis process above. - Facilitate prioritization and alignment among stakeholders to identify 2-3 workflows that should be analyzed in more detail. Consider impact, feasibility, and risk. - Write opportunity briefs that explain, for each workflow: what customers are trying to do, what the agent should handle, what backend capabilities will be required, and what evidence supports the brief. - For each brief, gather the context needed to make the workflow real: SOPs, scripts, policies, escalation rules, API documents, schemas, sample payloads, test data, and business value assumptions. Turn important policy requirements into testable acceptance criteria. - Define and build the backend the agent needs. This is usually a server that provides tools like: status, eligibility, search, book, escalate. Use test data so the demo behaves like a real system without needing production access. - Build the service agent configuration to leverage the backend tools: role, instructions, boundaries, tone, tool usage rules, escalation rules, and safety requirements. Keep it narrow and voice-friendly. The agent should only promise outcomes its tools are capable of. - Build a customer simulation agent based on scenarios seen in the original data. Have it interact with the service agent. Evaluate task completion, policy adherence, tool usage, and escalation. - Generate a Stage 1 proof pack that includes all documents, data, analysis, code, config, instructions, simulation and evaluation runs, results, and a data-derived value model and ROI analysis. Complete whenStage 1 is complete when the evidence is clear, the selected workflow is explicit, the backend capability runs, the agent package is active, the agent can complete or safely hand off representative cases, and the proof artifacts support a credible ROI conversation. Stage 2Secure the launch path10 steps - Start from the Stage 1 workflows. Use the same evidence, backend, service agent, simulation agent, and proof artifacts as the baseline. - Confirm the readiness bar. Define what must be true for performance, policy adherence, customer experience, escalation, infosec, compliance, and business value for the agent to be deployed into production. - Expand the simulation set. Add more functional scenarios, harder edge cases, and realistic failure paths based on the original conversations and the Stage 1 proof findings. - Add adversarial and red-team testing. Test for identity bypass, social engineering, prompt exfiltration, confidential data leakage, false confirmations, unauthorized actions, unsafe advice, and attempts to push the agent outside its scope. - Use evaluation results to improve the agent. Update prompts, tool descriptions, backend behavior, knowledge, escalation rules, and scope. Then rerun the relevant simulations to prove the fix worked. - Quantify performance and risk. Report task completion, safe handoff rate, automation potential, policy failure rate, attacker success rate, critical violations, tool errors, and any remaining gaps. - Decide how the new experience will integrate into the existing customer journey. Detail the integration points with existing systems and external dependencies that ensure the right contacts reach the new experience, and failed or partially completed interactions are handed off smoothly. - Refresh the business case. Use the stronger proof to estimate automatable volume, gross labor saving opportunity, volume, type and handling cost of escalations. Update platform cost assumptions and ROI based on the target platform and operating model. - Prepare a reusable simulation and evaluation harness that can be used to validate the production deployment. - Create a Stage 2 readiness pack. Include the current agent assets, backend tools, proof runs, transcripts, scorecards, risk and compliance summary, launch-readiness assessment, business value model, and a clear list of what remains out of scope. Complete whenStage 2 is complete when at least one of the Stage 1 workflows has survived broader functional testing, adversarial testing, policy testing, and value analysis; the remaining risks are explicit; and business, operations, technical, security, and compliance stakeholders have enough evidence to decide whether to move toward production. Occasionally, none of the Stage 1 workflows will meet that bar. Again, that is valuable information that avoids a failed deployment and means the organization can refocus its CX transformation efforts. DeployProduction deployment8 steps - Hand the Stage 2 repository containing agent instructions, config, code and stubbed backend implementation along with the reusable simulation and evaluation harness to the production deployment team. - If deploying with the same agentic AI tech as used in the CX Foundry process, simply deploy the Stage 2 assets into a production environment. For other environments, a short ‘translation’ pass is required to refactor the agent instructions, config, and code for the target environment. - Replace demo backends with real integrations. Map each test tool to the real API, database, queue, workflow system, or human escalation path. Define authentication, permissions, schemas, retries, idempotency, logging, and failure behavior. - Load test data into the backend systems that matches what was used in Stage 2. Re-run simulation and evaluation to validate the deployment. - Complete security, compliance, and operational review. Confirm data handling, access controls, secrets management, audit logs, human approval boundaries, retention policies, incident response, and rollback. - Define monitoring and support. Track task completion, handoff rate, tool errors, latency, cost, customer satisfaction, policy violations, escalation quality, and model or prompt regressions. - Run a staged rollout. Start with internal testing, then shadow mode, then a controlled pilot, then broader production traffic only when live results match the Stage 2 evidence. - Convert estimated ROI into measured ROI. Replace assumptions with production data: actual call volume, containment, average handle time impact, labor savings, platform cost, implementation cost, and net value. Complete whenProduction deployment is complete when the agent is fully deployed, integrated, handling a significant portion of live traffic, and performing close to or beyond the targets established during the simulation and evaluation activities in Stage 2. **CX Foundry Method: AI Implementation Guide** Paste this into a fresh chat, then ask an AI assistant to guide the work stage by stage. Copy Markdown ### Creative Commons Attribution-ShareAlike 4.0 International This method text is provided under a Creative Commons Attribution-ShareAlike 4.0 International license. You're free to copy, use, and adapt it, including commercially, as long as you credit CX Foundry, link to the license, indicate if you made changes, and share public adaptations under the same license. The license covers this method text, not CX Foundry trademarks, client data, or implementation code unless separately stated. [Read the license](https://creativecommons.org/licenses/by-sa/4.0/) --- 06**Get started ## Start with a Snapshot. If you're happy with [how we work](/how-we-work), and [how we look after your data](/data), and can share a sample of customer conversations, the CX Foundry Snapshot is a low-cost, low-commitment way to make meaningful progress in your GenAI journey. Just 1-2 weeks after the data usage terms are agreed, and you transfer the conversation data to the secure Azure blob storage location, we come back with an executive readout: what's happening in your conversations, the top GenAI opportunities, a custom interactive voice AI prototype that brings the opportunity to life, and a recommendation on what to do next. You get access to the analysis tool and custom voice demo for 90 days via the Client Portal. ### Submit a request To get started, share your name and work email address. We'll review the request and, if there's a fit, we'll reach out to confirm scope, timing, and commercial terms before the work begins. --- ## Privacy noticeLast updated: June 3, 2026. This notice explains how CX Foundry handles personal information on this website, in the Get Started form, and in the Client Portal. CX Foundry is a Waterfield Tech program. For the separate handling of client conversation samples, transcripts, recordings, engagement files, and AI delivery data, read [How we handle your data](/data). #### What we collect When you use the Get Started form, we collect: first name; last name; work email address; the date and time the request was received; the source of the request, so we know it came from the CX Foundry website. When you use the Client Portal, we may also process sign-in details, account status, portal access records, terms-acceptance records, engagement status, document metadata, and audit events needed to keep the portal safe and useful. When you browse the website, our hosting and security systems may receive ordinary technical data such as IP address, browser, device, requested page, timestamp, and error logs. #### Why we use it We use this information to: validate your request; send or prepare a Client Portal invite; protect the site and portal from misuse; maintain a record of consent, access, terms acceptance, and engagement activity; respond to your questions; improve the reliability and usefulness of the website and portal. We do not sell this information. We do not use it for cross-site behavioral advertising. #### AI-assisted handling We may use Waterfield-approved AI tools to help summarize, classify, de-duplicate, validate, or draft internal follow-up from an inquiry or portal record. AI tools do not make binding commercial, legal, access, or engagement decisions. A Waterfield human remains responsible for those decisions. We do not use your personal information to train or fine-tune a shared AI model. Where AI is used in CX Foundry delivery work after an engagement begins, the applicable Waterfield Tech terms govern that work. #### Who handles it for us For this website and portal intake flow, the main systems are: Google Cloud and Firebase for website hosting, portal authentication, Firestore records, application logs, security controls, and runtime infrastructure. Google Workspace for internal email, documents, calendar records, and human operating records. These systems may process information in the United States and other locations where they operate. Waterfield Tech personnel and approved advisors may access the information where they need it to operate CX Foundry, support the portal, respond to a request, or meet legal and security obligations. If we add a material new provider for this website or portal intake flow, we will update this notice. #### How long we keep it We keep Get Started form and portal-intake records for as long as needed to handle the request, maintain a reliable operating record, meet legal or security obligations, and support any resulting engagement. If no engagement starts, we periodically review and remove records that are no longer needed. #### Cookies and tracking The public website does not need advertising cookies to work. The Client Portal may use essential cookies or similar storage for sign-in, session security, and portal operation. Those are used to provide the service, not to sell your information or track you across unrelated sites. #### Your choices You can ask us to: confirm whether we hold personal information about you; correct information that is wrong; delete information where we no longer need to keep it; stop contacting you about a request. Some records may need to be retained where required for legal, security, contract, audit, or fraud-prevention reasons. #### Contact For privacy questions, contact Waterfield Tech at [support@waterfield.com](mailto:support@waterfield.com) and include "CX Foundry privacy" in the subject line. Mail: Waterfield Tech, 110 S Hartford Ave, Ste 2502, Tulsa, OK 74120, United States. #### Changes We may update this notice as the website, portal, processors, or operating model changes. The updated notice will be posted on this page with a new last-updated date. --- **CX Foundry**A Waterfield Tech program [How we work](/how-we-work)[Principals](/principals)[Method](/method)[Data](/data)[Privacy](/privacy) © 2026 Waterfield Tech