VP, CX Insights & Community at PartnerHero | Talking about creating your customer experience from scratch. I build and scale customer-facing teams that give a shit and can help you do the same.

I transitioned from being a barista to working in customer experience 20 years ago, and life hasn't been the same since.

I've been fortunate enough to work at companies like Wistia, Atlassian, Trello and Appcues. I've worked with amazing customer-facing folks to create scalable strategies and delight customers. 

Now, I spend my time helping to elevate leaders of teams and empower them to create the best customer experience. Employee-centered teams with simple, beautiful processes are where I find my joy.

Outside of my work with Boldr, I also consult and write for businesses such as Help Scout, HubSpot, and others. Let's work together!

Mercer Smith

Every so often I see someone say: “Customer support just doesn’t want to help.” And I am almost always like "wait wut."

No one wakes up thinking, “You know what would really make my day? Getting yelled at by a stranger on the internet.” or “I hope this escalates so I can do more emotional labor for the same pay.” There is absolutely zero incentive for a support rep to not help you. Helping you = resolution. Resolution = fewer follow-ups. Fewer follow-ups = a better day.

Do you really think a human would rather be trapped in a hostile, circular conversation than just solving the problem? What people interpret as “they don’t want to help” is usually just policies they didn’t write, tools they didn’t choose, rules they can’t override, and systems that make everything harder than it needs to be.

Most support people want to help, but might not have the tools or privileges required to do so. Beyond that, they literally just want to keep their job. It's not that deep and they don't have a personal vendetta against you.

CX Strategy & Solutions at Boldr; Author of CXOXO | Talking about creating your customer experience from scratch. I build and scale customer-facing teams that care and can help you do the same.

Pro-tip: if you are going to travel wearing your work gear (branded etc), maybe don't be a grossy.

Example: talking about how to make foreign people sound "less foreign" because "no one wants to talk to them."

If you were in the Manila Silver Kris lounge at 8.31am on Saturday, Jan 17th, this is for you: accent "translation" software is gross. The folks that "no one wants to talk to" usually speak like, probably (on average) two more languages than you do. You suck.

Imagine getting upset about something I wrote on LinkedIn, and meanwhile I’m over here typing with fingers that look like this 👀

No alternative text description for this image

new year, same me.  A bad CSAT score is not a reflection of your team members.

Customers use CSAT to say:
 – “Your policy is garbage.”
 – “Your product is jacked.”
 – “I had to reach out when I'd rather just be NOT doing that.”

Stop judging humans by metrics that measure company failures. Use it if you want, but be clear about what you're actually trying to measure.

Last night was the kind of dinner that makes a few hours disappear.

Boldr and Richpanel hosted a small, invite-only conversation on CX Transformation Through AI: Humanizing the 2026 Experience and it was exactly what we hoped for: candid, practical, and genuinely human. Huge thanks to everyone who showed up ready to share real stories (and real opinions), ask better questions, and build on each other’s thinking. Tagging ya: Alex, Kellie, Colin, Jason, Olivia, Ashley, Ashley M., Jordan, Ken, and Sid.

I pulled together a short report to accompany the night: a mix of fresh research and anonymized field notes from the conversation. If you’re thinking about how to scale CX with AI without losing the human connection, it’s worth a skim.

A few themes that stuck with me from the convos:

1. Customers want AI when it’s simple, and a human when it’s not. The “best” experiences weren’t fully automated. They were well-designed blends with a clear escape hatch.
2. Agent-side AI is where a lot of the real wins are right now. Summaries, drafting, translation, routing make for less cognitive load, better consistency, faster resolution.
3. Context handoffs are still broken far too often. If customers have to repeat themselves, the experience feels colder instantly no matter how “smart” the bot is or how kind the human is.
4. Human connection is becoming a premium differentiator. The stories people shared about handwritten notes, thoughtful follow-ups, and community moments were a reminder: loyalty is emotional.

If you're curious for more, here's the Gamma link to the report that I will send out to all of our attendees as well: https://lnkd.in/gYFkPfG6 If you read it, I’d love to hear what resonates (or what you disagree with).

And one more thing: I’d love to do more of these. If we brought a small, facilitated dinner conversation to your city, where should we go next? NYC feels like an obvious next stop, but tell us what you’d vote for and if you'd be interested in attending or hosting.

How all of my emails after two weeks of vacation are going to find me tomorrow.

Sixteen years ago, one of my first talks (shout out ElevateCX) was basically about how support people should ingratiate themselves to engineers so they’d help build the tools we needed to do our jobs, and I meant it sincerely. That was the move. You were a “lowly support person,” so you built relationships, baked cupcakes, asked nicely, and hoped your request didn’t get deprioritized.

Later, at Trello, Ben McCormack and I spent an absurd amount of time building janky terminal tools just to get basic visibility into things we needed. It worked, technically. But it was duct tape and vibes. 

Yesterday, I built a small app to fix a really silly manual process in my personal life. From idea to working thing took less than 2 hours. Here’s what the app does: forever, I’ve used a random number generator and a spreadsheet to pick the next book I read. It’s clunky, but it worked. Then, in the middle of the night, I finished a book and absolutely did not have it in me to manually count rows and roll dice like it was 2009. So instead, I built a tool.

It scrapes my Audible, Libro, Libby, and Kindle libraries, pulls all the books into a single list, and lets me press one button to randomly pick what I read next. While I was filming the demo video, I realized there was one more feature that would make my life easier. You can literally see me type it in and it’s implemented moments later.

That’s when it really clicked how outdated our conversations still are.
We keep talking about AI like it’s THE TOOL. Auto-replies. Deflection. Summaries. Fine. Cool. But that’s not the real shift: the real shift is that support DOESN'T HAVE TO WAIT ANY MORE.

As a support person today, you can build the thing you need. You don’t need an engineer’s blessing. You don’t need to lobby for priority. If you understand the problem—and support always does—you can create a solution that works for you. But no one is talking about that?

We’re still framing AI as something that happens TO support, instead of something support can use to build, fix, and unblock itself. The bottleneck isn’t engineering access anymore. It’s whether we’ve updated our mental model of what’s possible.

Support has always been closest to the pain. Now it’s close to the power, too.
If we’re not talking about that, we’re missing the point entirely.

A lot of people replied to my last post with some version of: “Yes, exactly. They want to help. They just don’t have the tools.” Cool, we agree. But like, so what?

If your team WANTS to help but keeps getting blamed for bad experiences, that’s not a motivation problem. We've established that’s a systems problem. and SPOILER: systems are fixable.

Question if your team can actually solve problems, or if they mostly spend their day explaining why they can’t. If most of their job is apologizing for policies, waiting on approvals, or escalating basic stuff, you didn’t build a support team, you built a human buffer.

Ask yourself how often the same issues keep coming back. If the same complaints, questions, and edge cases show up every week, that’s not customers being annoying. That’s a signal. Something upstream is broken, and your team is absorbing the emotional fallout. Quantify that and fix it.

Get curious about what you’re actually rewarding. If your metrics prioritize speed, deflection, or “closing” over real resolution, you’re training people to rush, not help.

Dive in to how much real authority your team has. If every refund, exception, or fix needs approval, you might think you're creating financial safety for your business, but you're actually just creating more backlog. You're making a bigger wall between your team and the customer.

If you want your team to help more, make it easier for them to help. Most support people already care (I say most because there were some folks on the last post that said that wasn't true, hahah). Your job as a leader is to stop making it so hard for that to matter.

Not work-related, but YOLO: My son is doing a science fair project looking at the intelligence of different types of cats (hairless versus non-hairless). If you have a cat and feel so inclined to fill this out, we would really appreciate it! Once per cat:

Open article: 🐾 How Smart Is My Cat? by docs.google.com

Are you trying to get in on this? It's gonna be pretty good! Let me know if you'd like an invite and I'll see what I can do...I might know a guy.

We're hosting an intimate, invite-only panel in Austin with our partners at Richpanel. 

Join Mercer Smith, and Olivia Parker as they explore how AI is reshaping customer experience without losing the human touch.

What to expect:

• Practical insights on designing human-centered CX enhanced by AI
• A look at what CX leaders are prioritizing for 2026
• A thoughtfully curated group of 20 DTC and tech leaders

And after the conversation, we’ll slow things down with great food, drinks! 

Drop an “I’m interested” in the comments to save your spot (limited seats).

📅 January 29, 2026
📍 Austin, TX (venue shared upon approval)
🎟️ Invite-only | 20 seats

Customers don’t think in channels. Your data shouldn’t either.

In this co-written blog with Richpanel we show how to build a minimum viable omnichannel data model that finally keeps context where it belongs. We also include a channel map template to help you get started.

For CX leaders who want to go deeper, we're co-hosting an event in Austin with Richpanel at the end of January focused on building continuity across human + AI interactions.

Hope to see you there 👀

Read the full article here: https://hubs.ly/Q03ZZYL50

Open article: One customer brain for 2026: building the minimum viable omnichannel data model by hubs.ly

One customer brain for 2026: building the minimum viable omnichannel data model

Have you heard me talk about how you need to change your presentations to be more executive-facing if you want to actually get buy in and budget? And have you simultaneously been like "yeah okay sure but how?" (I hope not because I always try to give examples and steps to follow along, but fine.)

WELL LUCKY FOR YOU Craig Stoss AND I ARE OVERACHIEVERS AND not only are we going to give you a framework, but we also have examples of before and after slides USING the framework.

Don't snooze, we're going live in a few hours: https://lnkd.in/gKk3ufiE

New year, new Masterclasses! We are back fresh and reloaded with new sessions for our CX leaders. 

Join us and KODIF this January 22nd for a masterclass on how to turn CX data into executive-ready stories that drive buy-in, trust, and investment.

This session is ideal for CX, operations, and customer support leaders who need to communicate impact clearly, influence decision-makers, and consistently win investment for their teams.

See you there! 👀

Tyler Sellhorn, Aitana Rangel Garibay and 32 other attendees

View event: Executive-ready CX storytelling

so stoked to be able to support Gruns. <3

It's a new year. Six months ago, I shared a list of software and service partners who have supported our staggering growth at Grüns. 

I want to spotlight some new partners and continued heavy hitters who are making an impact in this next chapter of growth.

None of this is sponsored. Our team is simply grateful for the individuals behind these platforms, who are putting in the work to support Grüns.

Skio - subscriptions
Postscript - sms
Haus - incrementality
Northbeam - MTA
Siena AI - empathic CX
Boldr - CX extension
NoFraud - fraud prevention
Fulfil - ERP
Loop - returns
Numeral - taxes
Gladly - CX
Crstl - EDI connection
onbrand.com - partner activations
NeonPixel - TV performance
ShipNova - supply chain
Hark - voice of the customer
Motion (Creative Analytics)
Replo - site build
Archive - social listening
Order Editing - post-purchase edits
Chargeflow - chargebacks
PLTFRM - retail
Raindrop - creative production
Baker Tilly - audit and tax
J.P. Morgan Disruptive Commerce - LOC & banking
Ramp - credit card & AP
WithCoverage - insurance
Yolodex - persona creation
Hightouch - reverse ETL
AthenaHQ - lifecycle personalization
FrontrowMD - medical badging
Tracksuit - brand awareness tracking
Hetal Retail - retail analytics
Great Northern - retail display

I hope these partners are helpful for your business as well!

Last week on the CXL peer group call, we ended up in a conversation that I SWEAR every support/CX leader has had in some form but somehow we all keep treating it like a personal failing instead of shitty biz practice.

How many times have you been told that to be considered for a promotion, you need to do the work of the new role (plus your existing role). Your calendar fills up and the “extra” work starts showing up without clear role evolution or compensation. It's "for your development."

One person said something that stuck with me: if someone wants you to step into a bigger lane without paying you for it yet, the only sane move is: “Cool. Help me understand how this role evolves over time.” Not “someday.” Not “we’ll see.” Literally: what changes in 6 months, what changes in 12 months, and how will success be evaluated. Because if the company can’t answer that, someone else in the market will.

And here’s the piece I feel annoyingly strongly about: don’t let yourself get into the pattern of doing the bigger job without being compensated for it. The first “sure, I can cover that” is the easiest moment to set a boundary. After 6–8 months of wearing the extra hats, it turns into debt. And businesses are gonna business; if they can juice you, they will. 
I liked the flipped perspective though, too: sometimes it is less about immediate comp and more about increasing your luck surface area. However, if you take on this “extra,” what are you no longer able to do that might be worth more?

We also got into a super useful comp angle that I don’t hear enough leaders talk about: if base salary is capped, get creative. Stipends tied to projects. Variable comp tied to outcomes. Commission if you’re helping with pre-sales. A defined time-boxed trial with an explicit renegotiation date. There are ways to avoid the “just keep doing it forever” trap that don’t require a dramatic ultimatum.

Anyway: I’m sharing because I think everyone talks about career growth and comp negotiation, but we don’t talk enough about the patterns that quietly trap support leaders: the title band-aids, the unpaid scope creep, the “you’re so capable” tax, the way helpful people accidentally train companies to underpay them.

If you’re navigating one of these right now, here are two questions from that call that I think are worth stealing:

1. What does this role look like 6–12 months from now, and what specifically changes (scope, decision rights, comp), if I execute well?
2. What am I agreeing to do temporarily that I might accidentally make permanent?

Stripe

- Part of Stripe's Developer Advisory Board
- Providing early feedback on features
- Advising on developer experience

Developer Experience Advisor

Postman

- Consulting with enterprise organizations on API strategy: lifecycle, platform, governance.
- Building relationships within the API industry: product vendors, customers, and community members.
- Publishing research both internally and to the broader API community.

API Strategy Lead, Open Technologies

- Working to help enterprises take full advantage of Postman's Collaboration Platform for API Development.

Lead Solutions Engineer

Kevin Swiber Consulting

- Strategy consulting for SaaS-based startups.
- Custom Kubernetes and Knative development and integration.
- Work with Knative Open Source project on autoscaling opportunities.

Owner

OpenAPI Initiative

Outreach Chair

NodeSource, LLC

- Successfully led the company through an acquisition, spinning down infrastructure under
the old company (NodeSource, Inc.) and spinning it back up under the new company
(NodeSource, LLC).
- Interacted with key customers to retain their business and closed over $1MM in revenue
in the company's first quarter.
- Influenced new product direction.
- Built Chef cookbook to help win a customer deployment.

Principal Software Engineer

NodeSource

- Led design, reliability, and observability of a new SaaS product.
- Jumped in to fill a need for an SRE. Managed resources on AWS using EC2, Lambda,
ECS, IAM, CodeBuild, CodePipeline, S3, RDS, SQS, CloudFormation, CloudWatch,
Elasticsearch Service, Route 53, CloudFront, and other services.
- Managed repositories of CloudFormation templates driven by Troposphere and other
resources driven by Terraform.
- Implemented and taught Prometheus best practices for metrics and alerting.
- Provided architecture and software development mentorship to team members.

Product Architect / Senior Software Engineer / Site Reliability Engineer

LunchBadger

- Developed, released, and maintained Express Gateway, an open source, policy-driven
API Gateway built on Express.js and Passport.js.
- Worked directly with Marketing, Sales, and Product to build the product roadmap,
develop features, create product announcements, and support pre-sales engagements.
- Led the Engineering team to create a Kubernetes-native platform for running API
Gateways on top of managed services.

Head Of Engineering

Apigee

Responsible for designing and developing research projects, released as Open Source Software, and providing community thought leadership around API design.  Built and launched Zetta IoT connectivity platform and Apigee Link IoT product.

Engineering Manager - IoT, Microgateway, Edge X

Quicken Loans

Responsible for cloud integrations and process improvement.

Enterprise Architect

Developed an industry-leading data-driven system for serving leads and managing appointments.

Software Engineer

EHIM

Outlined architecture standards and process improvement measures.

API Strategy Lead at Postman

I've worked with API-focused startups for the past decade. I teach and learn about software development, distributed systems, leadership, mental health, and developer tools.

With a strong focus on opening information channels via Web APIs, I work to help organizations improve their API Lifecycle.

Kevin Swiber

Your MCP tool descriptions are eating your context window.

I've been reviewing MCP implementations, and the same pattern keeps appearing: verbose tool schemas that burn thousands of tokens before the agent does any actual work.

One data point that should make everyone pause:
- 106 tools
- 207KB of schema data
- roughly 54,600 tokens
- on every initialization

Even when the model only needs 2 or 3 tools.

This hurts cost, latency, and tool selection accuracy at the same time.

The good news: this problem is getting serious attention at multiple layers of the stack.

🔍 The Three Layers

Token efficiency isn't a single problem. It's three problems at three architectural layers:

1. Server-side - How verbose are your tool schemas?
2. Protocol-level - Does MCP support lazy or filtered discovery?
3. Host-side - Does the MCP host send every tool to the LLM?

That last one is crucial and often overlooked. Your MCP host doesn't have to forward every discovered tool to the model. It can filter, search, or progressively disclose before anything hits the context window.

🔐 The Ecosystem Response

The market isn't waiting for protocol changes.

Claude Code just rolled out MCP Tool Search. When your MCP tool descriptions would consume more than 10% of context, tools are now loaded via search instead of preloaded. This addresses one of the most-requested GitHub features: users were documenting setups with 7+ servers consuming 67k+ tokens.

On the server side, MCP gateways are collapsing hundreds of tools into just "search" and "execute." On the host side, proxy layers intercept manifests to apply semantic filtering. Agent Skills demonstrate progressive disclosure already shipping across multiple agents.

The MCP community also has proposals in flight showing 91-98% token reduction is achievable at the protocol level.

🤝 Three Things to Remember

1. The MCP host layer is your lever, especially if you're building your own agent. Claude Code is already doing this with Tool Search. If you're building your own agent, you can implement similar filtering.

2. Every word should earn its place. Concise descriptions, semantic naming, and clear server instructions help both token budgets and discovery accuracy.

3. You're not alone! Gateways, proxies, and patterns like Agent Skills already exist. If you're hitting context limits, someone in the ecosystem has likely solved your variant.

I wrote up the full details with links to all the GitHub proposals, code examples, and implementation patterns here: https://lnkd.in/g9nB_s9H

What's your approach to managing tool manifest size? Are you filtering at the host, using a gateway, or waiting for protocol changes?

#MCP #TokenEfficiency #AIAgents #ProtocolDesign #DeveloperTools

API & MCP Strategy | Ex-Postman, Ex-Apigee

Open article: MCP Tool Schema Bloat: The Hidden Token Tax (and How to Fix It) | Notion by layered.dev

MCP Tool Schema Bloat: The Hidden Token Tax (and How to Fix It) | Notion

taplio

The Battle

Mercer Smith

Kevin Swiber

Followers

Posts

Engagements

Likes

Performance ScoreMercer Smith's LinkedIn Performance

Performance ScoreKevin Swiber's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Mercer Smith

Kevin Swiber

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Mercer Smith's LinkedIn Performance

Performance ScoreKevin Swiber's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreMercer Smith's LinkedIn Performance