The questions
people actually ask.

Every engagement starts with a free 30-minute intro call to confirm fit. If we move forward, Intelligent CX Consulting scopes a 1–3 week discovery and delivers a proposal within a week of that call. First artifact — a documented friction map, an intent taxonomy, or an eval harness, depending on where you're starting — ships inside the first month. No mystery, no open-ended retainers, no billable hours for meetings that should have been a Loom.

I run every engagement through what I call Thoughtful Execution: identify multiple friction points, test different approaches against real data, and ship what actually moves metrics. In practice that means two weeks of watching how work really flows before I propose a single change, followed by small, reversible experiments with clear rollback criteria. I don't trust "we should probably do X." I trust "here's the data that says X works."

The same way your users do: did it work? I care about containment, latency, satisfaction, and cost, not vanity metrics. If the number moves and nothing real changed for the user, I treat that as a warning, not a win. I also track something most teams don't: whether the system keeps working after I leave. Rituals, documents, and handoff specs survive me, or I didn't do my job.

Measurable eval improvements on the flows that actually move your north-star metric. A documented intent taxonomy your product team can reason about without me in the room. An internal playbook — prompts as versioned specs, eval rubrics, escalation rules — your designers and engineers can run without another consultant. If six months in your team still needs me to be the smartest person on the call about your own AI, I failed.

Almost always with a team. My job is to make your existing engineers, data scientists, designers, and product leads more effective, not to replace them. I specialize in the translation layer between roles that tend to speak past each other.

I don't fight engineering constraints. I treat them as design opportunities. "We can't do that" almost never means the user problem is unsolvable, it means the specific approach we brought isn't the one that fits the constraint. I find a different one within bounds.

I do foundational coding. I'm comfortable in JavaScript and Python, and I pair heavily with Claude (Anthropic) and other AI tools to move faster without losing intent. I'm not here to replace your engineers, but I'll never be the person who can't read what your team is shipping.

Three reasons. One: Claude's writing is better at restraint than most alternatives, which matters when you're designing for customer support and trust. Two: the tool-use and multi-turn behavior holds up under production pressure in ways I've independently verified. Three: Anthropic publishes research I actually use in my work (AI Fluency, constitutional AI, evaluations), so the ecosystem tracks what I care about. That said, I ship on whatever stack the team already runs. I've deployed on Dialogflow CX, Azure AI Foundry, and AWS Bedrock. The model matters less than the orchestration around it.

Multi-agent orchestration (role-specialized agents with a thin orchestrator), prompt architecture as versioned specs (not prompt strings in production), LangFuse for observability, and eval harnesses built on day one, not bolted on later. Day-to-day I live in Claude (Sonnet and Opus), Cursor, Figma, Dialogflow CX, Azure AI Foundry, and AWS Bedrock. I use Notion for specs, GitHub for everything else.

Claude (Sonnet for most work, Opus when the reasoning matters), Cursor for pairing, Figma for flows, LangFuse for traces and evals, Astro for this site. On client stacks I work in Dialogflow CX, Azure AI Foundry, AWS Bedrock, and whatever prompt-orchestration layer the team has already committed to. The tool matters less than the evaluation harness around it.

The agent hype is mostly a category error. What people call "agents" are usually orchestrated workflows with LLM-backed decision points, which is useful, but the word "agent" is doing a lot of heavy lifting. The real unlock isn't agency. It's orchestration: AI doing the cross-system work humans were manually doing. That's the case study I'm most willing to die on a hill about.

I'm a natural conversationalist. I enjoy the spoken and written word, always have. Coming from hospitality, I learned that every word is intentional, that reading the room is a technical skill, and that the difference between "understood" and "felt understood" is where real experience lives. Conversational AI is that exact problem at scale. The interface is the writing. The writing is the product. I don't have to choose between my favorite things.

GED → non-traditional student + hospitality → workflow optimization → conversational AI. My path taught me what most in my field miss: not everyone thinks alike. I learned to communicate by working with people who had different vocabularies, different tech comfort levels, different needs. That's my entire job now. Bridge people who don't share language. Make systems that serve global users. Plain language. Accessibility. No exceptions.

Teams who've already shipped something, know where it's falling short, and want a partner who will tell them the truth about why. I'm less useful to teams who want a vendor to rubber-stamp what they've already decided.

Projects that exist to launch a press release, not to serve a user. Companies shipping AI they'd be embarrassed to show to their grandmother. I've done the work of building for real people at scale; I'm not going to unbuild it on a side project.

Best fit for 3–12 month retainers where we can ship, measure, and iterate. I also take 2–4 week diagnostic sprints when a team needs an outside read before committing to a larger build. I don't take multi-year vendor contracts; by year two you should have internalized the playbook.

Yes. Half-day and two-day formats on eval-driven conversation design, intent taxonomy, prompt architecture as product spec, and multi-agent orchestration. Audiences range from product and design teams to ops leaders who need to understand what AI can and can't do for their workflow. Email for topics and availability.

I don't take open-ended 1:1 mentorships — my calendar won't support it. I do host occasional paid portfolio reviews for conversation designers transitioning into AI product work, usually in cohorts. Meanwhile, the ICX blog and the essays on this site cover most of the ground I'd repeat on a call.

Sometimes. The bar is whether I can point at a user problem worth solving. If you have that, I can help shape what to build. If you don't, I'll tell you that before I take your money, and point you toward finding it.

Book a free 30-minute intro call from the Contact page, or email [email protected] with a one-paragraph brief. Engagements run through Intelligent CX Consulting, LLC at https://www.intelligentcxconsulting.com. Either path lands on my desk and I respond within two business days.

Dallas–Fort Worth, Texas. I work remote with US clients by default — the Texas Triangle covers most of my in-person work. I travel quarterly or by request for workshops, on-site design sprints, and conference talks. International engagements are case-by-case, typically as remote advisory with one or two in-person weeks.

Retainer by default for multi-month engagements. Fixed-scope pricing for 2–4 week diagnostic sprints. Hourly only for expert-witness work or board-level advisory where the time is genuinely unpredictable. Specific numbers over email — I don't negotiate rates in public.

Everything on /work has been scrubbed or approved for public reference. Specific client attributions are available on request and usually fine once we've signed mutual NDAs. I won't share confidential metrics, internal documents, or team names without written permission — even on calls with prospects. I expect the same courtesy for your project.

Yes. I partner with a small set of design studios and data-science consultancies where a conversational-AI specialist is missing from their bench. White-label is fine if the end client knows a specialist is in the room and I can speak to them directly during scoping. I don't take work where the agency needs me to pretend I'm full-time staff.

Yes, and I'd love to be able to write about it too. My default is anonymized publication after launch unless we've agreed otherwise during scoping. I won't use your logo or name without written sign-off. If the work is under a strict NDA, I still want a one-line stat I can reference ("saved a Fortune 500 hospitality brand $2M in support costs") — we'll agree on the phrasing before anything goes live.

Yes to all three. I speak on conversational AI, orchestration, content engineering, and eval-driven product work. I guest on podcasts, write here and on the ICX blog, and take quarterly advisory roles where a team needs a named expert in the room. Drop me a note with date, audience, format, and budget and I'll reply within a week.