When Bret Taylor — co-creator of Google Maps, Chairman of OpenAI and former co-CEO of Salesforce — asks you to join his new start-up, the answer is yes. And that’s how Logan Randolph ended up as Sierra’s first go-to-market hire in 2023. “The advice I’ve always followed is: find the smartest people who take you seriously, and do whatever they ask,” says Randolph.
Sierra helps businesses build better, more human customer experiences with AI. Two years later, and 18 months since launch, the startup is used by hundreds of brands, including some of the largest enterprises in the world. Almost 50% of the companies using Sierra have annual revenue of over $1B, and 20% have annual revenue of $10B+. How did they attract such large enterprises so quickly? In short, first principles thinking combined with a design partner program to hone their product market fit.
Inspired by the launch of ChatGPT in November 2022, Sierra’s founders, Taylor and Clay Bavor (a product leader at Google for 18 years), made an early bet on AI agents.
"They had this theory that agents were to AI, what apps were to mobile and websites to the Internet. And just like websites and apps before them, agents would transform the customer experience,” says Randolph. “What wasn’t clear were the specifics.”
To validate their theses, Taylor and Bavor tasked Randolph with building a design partner program. Not the glorified beta testing most enterprise startups use, but real partners who were excited to build alongside Sierra. Over two years in, Randolph says he learned three important lessons about building a design partner program:
- Target recognizable brands from day one — companies with real businesses who can pressure-test every facet of your product.
- Bring your partners in early so you can build the product together — not at the end when everything’s almost done.
- Ensure both sides have skin in the game — with agreements that look like paid enterprise sales contracts not early-stage pilots.
Completely orient your company around your partners’ success — ultimately their success creates your success.
In this interview, he takes us back to the beginning, unpacking the playbook Sierra used to identify design partners, and go from prototype to product-market fit. Let’s dive in.
Finding the right partners: Define your criteria
Sierra didn't yet have an ideal customer profile — that's some of what the design partner program would eventually validate, so they cast a wide net. With the goal of landing four, the team came up with a list of 40 potential partners (some friends, some strangers). Here’s the criteria for how they selected these companies:
Criteria #1: Horizontal appeal
Many successful start-ups have gotten early traction selling to other tech companies. And it likely would’ve been easy for this founding team — with decades of connections as founders and executives across tech giants — to follow that same playbook.
“But we believed AI would transform the customer experience across every industry, and we wanted to validate that theory. So we went to people in healthcare, CPG, media, retail and tech without knowing yet which ones it would actually resonate with,” says Randolph.
Criteria #2: Large scale
Sierra’s early story is another reminder of how quickly AI adoption has accelerated. “We started with the pitch: You should use AI for customer service — which was actually non-obvious a few years ago,” says Randolph. “We wanted design partners that had to consider the risks associated with AI seriously, as we learned a lot from mitigating these risks together.”
There was also a direct correlation between how big the company was and the scale of the problem. “If you have two CX managers and you ramp up to four during the holidays, that’s not a huge deal. But if you have a call center of 5,000 employees and have to ramp to 7,000, that’s really tough to manage,” he says.
Criteria #3: Real problems
No matter the size of the company, Sierra wanted to avoid AI tourism. “In some conversations it became clear that prospective partners were interested in ‘playing with AI’ and building internal prototypes,” said Randolph. “We wanted customers who wanted to solve a real problem that meaningfully impacted their business.”
A fun experiment often starts with “wouldn’t it be cool,” and in the design partner stage, that was almost always a red light.
A real problem sounds like, “Every holiday season our agents have to work 24/7. It’s stressful for them and a bad experience for our customers. How can we use AI to scale our support, especially at the busiest times?”
Criteria #4: Not too many friendlies
Sierra had an advantage in that their founders had huge networks. But finding design partners who weren’t warm intros was important to validating Sierra’s thesis.
"It’s one thing to have close friends tell you it’s a great idea. But the views of strangers are not clouded by your existing relationship. So they hold you to a different standard," Randolph says.
Approaching partners: Beware the frictionless gap
"A lot of startups will say to prospective partners, 'We just need a few hours of your time each month.’ But you need mutual investment and risk,” says Randolph.
"So we told partners upfront: ‘We'll give you dedicated engineers, direct access to our founders, and our cell phone numbers. But in return, we need real investment from you — payment, access to your systems, and weekly meetings to get candid feedback,’" he says.
To get to locked-in, mutually invested design partners, Randolph led a four-step process:
- Step 1: Customer discovery + deck (30 minutes). “This was the initial meeting to judge mutual interest,” he says “To do this effectively, we shared context and asked a lot of questions, but we really weren’t giving the hard sell.”
- Step 2: Live demo (30 minutes). “We actually showed prospective partners our product to illustrate what we wanted to do together.”
- Step 3: Deep dives as needed (security/technical diligence). “Step three had more variance,” Randolph says. “At this point we followed the customer, addressing their questions and concerns. And at times going back to steps one and two to get buy-in from new stakeholders.”
- Step 4: Verbal commitment to the paid design partner program.
Step 1 is often overlooked by startups, eager to get someone (anyone) to take a chance on them. “At this point, we’re not even selling people on being a design partner,” says Randolph. But this process of real discovery to understand what potential partners would like from an AI agent is so important. Without it, you risk becoming a consulting firm, crafting bespoke products with wildly different directions for each partner.
So use this discovery call (and the demo and deep dive that follow) to get extremely granular on what success looks like for the partner. “Once we had the same goal — for example an AI agent with high resolution and CSAT rates — it was pretty clear what was in service of it, and what wasn’t,” he says.
If it’s within that scope, anything is fair game. “Need to have an NPS pop-up immediately after every conversation? Great, we’ll build that feature. Need to have a credit card feature so you can pay in the agent conversation? Great, we’ll build that. Need to have the ability to look at every conversation that’s happening? Great, we’ll build that feature,” says Randolph
“This is why we have so much gratitude for our design partners,” he says. “Every feature that differentiates our product from the market was an ask from one of our design partners.”
Partner’s goals need to be similar enough that you’re OK helping them achieve goals in unique ways. We didn’t take on design partners who we didn’t want to follow.
At the end of the four-step process, Sierra ended up with six design partners (against their goal of four), all of whom were also paying customers. Among them are:
Structuring the partnership: Mutually assured construction
When you’re undergoing a design partnership, with frequent touchpoints and heaps of feedback, you’re asking for a large commitment from your partners.
Randolph leaned into this model, with a clear end date and payment upfront. “Both the timeboxed partnership window and the payment were commitments for the partner. They had skin in the game,” he says.
Randolph suggests investing in a structured contractual approach for the POC. Even if it takes more time initially it will allow your partners to move to long-term customers more easily.
Here’s why he endorses this approach.
Payments
Putting up hurdles for early customers to jump through sounds like the opposite of Silicon Valley’s “move fast” ethos — but in Sierra’s case, it helped them disqualify companies they felt were AI tourists. “Everyone’s excited to experiment with AI. So the financial commitment had to be significant enough that people really needed to think about it, get approval from their boss and go through the procurement process,” says Randolph.
For pricing out your own design partnership, Randolph recommends a simple calculus. “10-20% of your total contract value feels right. Anything less than that and it’s not that real of a commitment,” he says.
Time limits
“Whatever you do, don’t allow the ‘try this and give us feedback when you are ready’ approach because they will say ‘Ok cool, we’re excited to work together, and we’ll get you resources next month,’” Randolph says.
Instead, he suggests setting tight guardrails: “If you say, ‘We’re going to work together for three months and then the partnership period is over,’ they’ll show up the next week ready to get to work.”
While each product and partner has unique needs, Randolph offers a helpful ballpark. “Under two months is probably too short and over six months is probably too long,” he says. “It needs to be long enough to build together, see results and close a deal — and short enough that there’s urgency on both sides to get going and stay engaged.”
“Today we’ve launched agents in a week,” Randolph says. “But when everything was new, it took a bit longer.”
Running the program: Build trust through rapid iteration
The contract is signed, but the partnership is just beginning. It’s important to be as intentional about the kickoff process as the selection and contract. This is a critical inflection point to clarify areas of ownership and decide what’s most important to build.
Kickoff meeting
Randolph recommends starting with a service engineering discovery session, a 60-90-minute meeting with the whole team. He had four main goals coming out of each kick-off meeting:
- Agree on the three or four main problems to solve. One company wanted an AI agent to do returns, exchanges and package tracking. For another, it was having the agent solve connectivity issues.
- Define swimlanes. “We needed clear accountability for who owns the customer experience and who owns the technical integrations,” he says.
- Commit to a real launch during the partnership, not a prototype. Randolph suggests getting a commitment upfront from the customer that this isn't just something for them to tinker with.
- Create ongoing touchpoints to avoid “out of sight, out of mind.” Randolph made sure to establish a standing meeting and a Slack or Teams channel to open a constant line of communication with the partner.
For Sierra, “more than half of the company was in every meeting,” says Randolph. On the partner side, after some trial and error, it was clear three folks needed to be in the room:
- The executive sponsor
- Someone on the technical side who could grant Sierra access to the APIs and connection points
- Someone on the business side who deeply understands the customer
We wanted to avoid starting to work with a partner and a month later they say, ‘Oh, we need to pull in this person.’ The kickoff was a forcing function to get everyone in the room and articulate who’s owning what.
After each kickoff, Randolph would send a recap email double-confirming that they’re all on the same page, as well as a short questionnaire: “With the questionnaire, I’m looking for anything that could sneak up on us later,” says Randolph.
Weekly iteration and communication
“We would try to get the first version of the agent built within two weeks. Even if it was bad, it was something,” said Randolph. These minimum viable agents were quite simple — for example, if the company wanted the agent to do exchanges, the MVP might just answer questions about exchanges from the help center.
And from there, progress was rapid. “Week over week, we were adding more fidelity. We began to operationalize and use our software development kit to build agents that did specific tasks for the partner, like completing a return, an exchange, or a warranty,” says Randolph. “And then, instead of using fake APIs, we’d use APIs that were connected to a sandbox, and then use production APIs that weren't launched.”
During the weekly 30-minute standups with design partners, Sierra put this progress front and center. “Every week we would explain what we did, we would solicit feedback and then we would take the feedback,” he says.
It was also an accountability check-in. “We would put up slides that spelled out clearly: Here’s everything we did, here’s the feedback you gave, here’s everything that we asked you for and you didn’t do yet. It was managing the relationship in a non-patronizing way.”
Prioritizing what to build
When it comes to design partners, choppy waters can throw your boat off course. How do you resolve the needs of multiple VIP customers when they don’t always align? It sounds like an enormously complex problem. But Randolph keeps it simple: “Anchor everything on the core problem. If the design partner’s request goes towards that core problem, then we’ll do it. If it’s not, then we won’t,” he says.
To build trust — even for some low-priority or hyper-specific requests — the answer was often yes. “We’d rather over-rotate on building things that are too specific as opposed to quickly disqualifying customer asks,” Randolph says. “Customers are usually right and while their asks may look different, they’re often aimed at solving the same underlying issue.”
But if you do find that you’re continuously butting up against requests that seem far afield, it probably traces all the way back to the selection process. “We had one early customer who kept changing what they wanted. Ultimately, this was a symptom of us picking a partner who didn’t feel the pain as acutely as other people. They fell into the bucket of ‘it’s cool to use AI agents, so let’s find a problem that would allow us to build one,’” he says.
And while the Sierra folks were eager to hop onto just about any problem, there were a few larger requests that were out of scope for the partnership window. “Our product started with just chat, but we kept hearing that people wanted voice, which wasn’t feasible at the start. But eventually, as the product matured, it was,” says Randolph.
“SiriusXM was one design partner who really pushed us on voice, and they eventually became our design partner for the voice product as well,” he says. “We were able to launch our voice agent into production before anyone else in the market, and we owe this head start to our partners at SiriusXM.”
Listening to really opinionated outlier requests has made our product better.
More than half of Sierra’s product offering today is directly attributed to design partner requests. “Part of the product is the agent itself, which interacts with customers. The second part is what we call the Agent Studio. You want to be able to read the conversations, look for trends on customer complaints and give feedback on what’s a good versus a bad answer so you can continue to improve the agent,” says Randolph.
But back in the design partnership days, the Agent Studio didn’t exist. And the feedback process left a lot to be desired. “We had links to conversations that were exported in a CSV. The design partner could then write up feedback and put it back in a CSV,” he says.
“SiriusXM really held our feet to the fire on this in terms of being able to give more specific feedback and see comprehensively what’s happening in each conversation, so we built Agent Studio,” says Randolph.
Rally internally around a milestone
Behind the scenes at Sierra, they were ruthlessly obsessed with one particular milestone: “We would pick a launch date when the agent would be deployed,” Randolph says. To get there, the team working on design partnerships had internal weekly syncs where they’d continually ask three questions:
- What do we need?
- What do they need?
- Are we on track with the launch date?
“The whole company was obsessed with these launch dates. Every single person could tell you that X customer was launching on May 1. It was the internal milestone that drove us,” he says.
Eventually, 100% of design partners would convert to customers — which Randolph credits to the team's maniacal focus. "If we hadn't gotten so specific, more customers would've driven toward a prototype. But because we were planning to launch agents, they held us to a higher standard," he says.
Wrapping up: The path less trodden
Running an impactful design partner program is extremely difficult — teams are on call at every moment, desperately trying to keep up with partner requests, a treadmill that never seems to stop. But it’s also a period of intense closeness to customers that has a certain magic at the earliest stages. “It was a shared experience that built customer empathy across Sierra. After engineers left their dedicated design partner post, they would look back and think fondly of that time working so closely with them,” says Randolph.
And the proof in Sierra’s approach goes much deeper than a 100% partner-to-customer conversion. “Champions at our customers have gotten promoted because we solved a problem for them that was so core to their business. We’ve had multiple design partners mention us on their earnings calls. They’re still among our most trusted collaborators and critical to the success and authenticity of Sierra,” says Randolph.
So if you’re going to do a design partnership, do it the right way, not the easy way. Your product will be all the better for it. “A true design partner program is not about early selling, it’s about building with your customers for the long term,” he says.