Director, Product Management, IBM Watson Commerce (2016 to 2018)
The next chapter began as a text-message exchange, and the only word anyone remembers from it was mine. In the spring of 2016, Richard Hearn, the partner who had said yes to experience-driven commerce back in my Adobe years, messaged me something to the effect of: I’m joining IBM, and you are too. I sent back the one word the entire industry would have sent. “IBM?” That one-word reaction became the opening of the blog post he published that July, under a title I could not have written for him: “The best startup I found. IBM?!” When I got him on the phone I asked, the way you can only ask an old friend, whether he was sober and what exactly was going through his mind.
He was serious, and his pitch was not about the company. It was about the people the company was importing. IBM had brought in Harriet Green, the executive who had turned around Thomas Cook, to run a business spanning Watson IoT and commerce, and she was recruiting outsiders with a mandate to shake the place awake. IBM knew what the market knew: it had spent a decade buying some of the best assets in commerce and marketing, Coremetrics, Tealeaf, Unica, Silverpop, DemandTec, Sterling, and then let too many of them atrophy while WebSphere Commerce, a platform born in 1996 with Atlanta Olympics ticketing in its origin story, aged in place. Richard’s line was that IBM finally understood this and was serious about the resurrection, and that I was going to come help revitalize the old acquisitions and the flagship. I had spent my Adobe years treating that flagship as the incumbent to beat, and once, memorably, as a quiet partner. Now the company on the other side of that strategy was recruiting me to run it. I was not buying.
Richard did not let it go, and the truth is I did not want him to; whatever my head said about IBM, the whole thing had an energy I liked. He was not selling me a job so much as recruiting a friend along on an adventure, and when he could not close me, he reached for the one introduction he knew I could not turn down: what if Michelle Peluso made the pitch herself? I had watched Gilt Groupe with admiration through my Adobe years. Its CEO had founded Site59 and run Travelocity, and she was flat-out a celebrity to me. I said I would love that, quietly certain it would never happen. It happened in about a day. Richard’s introduction landed in my inbox while I was in San Diego keynoting Tealium’s Digital Velocity conference, and Michelle’s reply came back that same evening: “Richard raves about you, and I’d love to chat.” The next day, between sessions, I took her call in my hotel room, pacing the floor and worrying the cell connection would drop. She asked about the market, the competition, and what I really thought of IBM’s assets, and I answered the way a Gartner analyst would: the acquisitions had likely already lost their windows, the talent had drained, the R&D had thinned. Then I told her, a little sheepishly, that I was not interested in joining but was glad to share what I knew. Two days later an email arrived from IBM’s HR Director for Commerce: “I understand you had a chance to talk with Michelle Peluso this week.”
A week after that I was at a bar in Berkeley with Kareem Yusuf, the senior executive who owned product management, engineering, and operations for all of IBM Commerce, and who would become one of the great mentors of my career. He had been pressed from somewhere above to meet me and told very little about why, so the first minutes had the feel of two poker players comparing hands they were not sure they were allowed to show. What was actually said over that first round stays between us. What I can tell you is what I walked out with: there was no open position, he would have to make one, there would be real engineering money behind the revival, and, above all, there was the way he talked about his own career. “This is what I do, I build it. And I deliver.” That night I sat with the question from the only angle that matters, why not, and could not find a blocking answer. The next day I emailed him a sentence I did not have to compose, because it was already true: “I have unfinished business with commerce. I would like to finish that work with you.”
Then I did my homework, and my homework told me to run. I spent a week working my network, and the verdicts came back with remarkable consistency. A friend who analyzed this market for a living told me IBM was “trying to take a rocking chair and turn it into a speedboat.” Word came back from an executive at Magento, through a friend there: “It’s a sinking ship that has taken on so much water it’s questionable if it can be saved. It’s a salvage job, not a resurrection.” Another friend diagnosed the structural trap in one sentence: the marquee customers had customized the platform so heavily that they could not absorb new releases, so every upgrade was effectively a re-implementation, which meant every upgrade was a competitive re-bid. And I had a safer path in hand. I was deep in interviews at Salesforce for a senior alliances role in retail, at a juggernaut in my own backyard, a company I had courted more than once. The same week my advisors were talking me out of IBM, I was still taking Salesforce screening calls. But the two paths were not really the same kind of thing. Salesforce was offering me a seat built around someone else’s platform. IBM was offering me the platform, the whole neglected estate of it, plus the people who had just spent a month convincing me to come rebuild it with them. The safe choice stayed safe, and the allure of the other thing grew into something almost mythical. IBM got to paper first, I walked away from the Salesforce path, and on July 11, 2016, I badged in.
The United Nations of software
I took over product management for two portfolios out of the gate. The first was Commerce: WebSphere Commerce itself, plus its troubled hosted edition. The second was Merchandising: the DemandTec pricing and promotion science I had been specifically recruited to run. Within months a third landed on my plate, Customer Experience Analytics: Coremetrics and Tealeaf. Together they came to nearly $400 million in revenue and a product organization of roughly thirty, working in lockstep with the engineering and SaaS operations organizations. My title contained a word the industry kept tripping over: I was a Director of Offering Management, because IBM had rebuilt its product discipline around the idea that a product manager should run the offering like a general manager runs a business. A retail trade publication that profiled me a year in put it kindly, introducing me as “the former Adobe and Intel exec with the curious sounding title”. The instinct behind the title was right and the word was wrong, and I spent my tenure translating it back to “product management” for everyone outside the building. Years after I left, IBM did the same, and retired the term.
Nothing about the scale was translatable. Working at IBM was like working for the United Nations: enormous, procedural, multilingual in the organizational sense, and vastly more challenging than I had imagined from outside. My first discovery was that I could not see my own business. Revenue arrived sliced into categories that took a finance degree to reconcile, spread across systems that did not talk, and there was no dashboard where a leader could simply look at the whole thing. So I built my own ledger in my journal and kept it current by hand for my whole tenure, a director responsible for hundreds of millions in revenue, doing arithmetic like a shopkeeper. That habit followed me out the door, not because every company is that opaque, but because the leader who has done the arithmetic personally argues differently in the room.
The second discovery was how far the flagship had drifted. WebSphere Commerce showed its age in ways both cosmetic and structural: the business tooling had run on Flash until late 2015, and when I looked at the complete history of what had shipped across versions 7 and 8, roughly seven years of releases fit on six sparse slides. The analysts were not subtle about it; my own strategy decks quoted the market calling us old, ugly and fat. What made this maddening rather than merely sad was that the product underneath was better than its reputation. When I got intimate with the platform, I found something the market refused to believe: against SAP hybris, the competitor everyone assumed had lapped us in B2B, a requirements-level comparison we built with a major systems integrator came out even, 64 requirements met apiece out of 112. A checklist is not a win rate, and hybris was beating us where checklists cannot see, on momentum, ecosystem, and story. But it told me the gap was not capability, and that changed what kind of problem I thought I had. We were losing on packaging, on economics, and on the accumulated distrust of a customer base that had stopped expecting anything new.
And then there was the hosted edition, Commerce on Cloud, which is where the distrust was being actively manufactured. The operational reality had fallen badly behind what had been sold: provisioning a customer environment was a weeks-long artisanal project, stability escalations were routine, and the unit economics were upside down, because every customer ran on expensively over-provisioned dedicated infrastructure. My first pricing analysis of it came back from my boss covered in gold text boxes, and one of them asked the only question that mattered: are we actually taking out cost, or just accepting a lower margin? I did not have a good answer yet. Getting to one became the education that every price book I have touched since is built on.
Ninety-eight business days
The low point comes before the good parts, because the good parts do not make sense without it.
By late November 2016, ninety-eight business days in, I had a typed list of ten impediments and a decision to make. Hiring freezes had landed on our business unit early in my tenure, freezing the very rebuilding I had been recruited for. The transformation mandate I kept being promised existed in conversations and nowhere else. The operational dashboards were worthless, the process overhead was crushing, and the scale that had seemed thrilling from outside felt, from inside, like weather: vast, impersonal, and beyond anyone’s control. I printed the list and asked Kareem to lunch. It is worth pausing on what that meant. Kareem was by then the executive over all of product, engineering, and operations for our business; my boss reported to him. A first-year director does not ordinarily get that table, and I had it only because of the way I had come in. We met at a Chinese restaurant in Berkeley, the same town where we had first met, and I slid the list across the table. He read it and figured out quickly what the meeting probably was.
What he did next taught me more about leadership than most books have. He did not spin me. He went down the list and was plain about which items were above even him, and which ones we could attack together. He challenged me to keep marching, kindly, and said he would understand if I chose not to. And somewhere in there he invoked the wild ducks, the way he did more than once in my time there. I looked the phrase up on the intranet afterward and found a half century of company lore underneath it: Thomas Watson Jr. had warned in 1963 that you can make wild ducks tame, but you can never make tame ducks wild again, and that IBM should never tame the ones it was lucky enough to attract. By my era it had become shorthand for the unconventional hires the company knew it needed and knew it might smother. I did not resign that day. I left the restaurant recharged, went back to work, and by spring the mandate I had been waiting for finally existed in writing.
I will not pretend the frustrations ended there. The reductions in force that swept through my team, one wave already behind us that first fall and a harder one the following spring, are the hardest thing in this chapter. Every one of those conversations tore at me, and I will not dress them up with a lesson; they were the cost of the year, and the people who left deserved better than the macroeconomics that took them. But from that lunch forward I stopped waiting for the company to become something it was not, and started working the list.
The freeze finally cracked in my second fall, and what it bought me was two product managers straight out of college, by way of a product-management bootcamp, offered on exactly those terms: take them or don’t. I will own my first read: after a year of fighting to hire commerce leaders from industry, it felt like a consolation prize. I asked the two of them to learn our products the way a sales engineer would and demo back to me. Within a month they were demoing live, standing up a working storefront from a prospect’s own product feed, and dismantling every low expectation I had walked in with. That pair changed how I judge potential.
Fixing the economics first
The turnaround had to start with money, because nothing else was fundable until the hosted business stopped bleeding.
The gold-text-box question had an empirical answer, and my team and I went and got it. We rebuilt the pricing analysis from the ground up: real average order values instead of guesses, apples-to-apples annual costs against the revenue-share pricing of the SaaS competitors eating us from below, honest margins computed on our own revenue instead of flattering ones computed on the customer’s total spend. The corrected picture was clarifying. Below a certain size of online business, our pricing model could not compete and should not pretend to; above it, we won more than the market realized. So we made the calls the analysis demanded, and they generalize well beyond commerce: know precisely where your model wins, refuse the business where it does not, attack the cost base hard enough to earn the right to expand later, and stop approving deals whose only virtue is the press release.
Then we performed surgery on the operating model itself. Inside IBM there was a world-class managed-hosting organization, acquired from AT&T, that had spent seventeen years running commerce operations for some of the biggest retailers in the country. My people were duct-taping customer environments at all hours while, elsewhere in the same company, sat professionals who did exactly that for a living (well, maybe not the duct-taping part). So we negotiated a formal agreement between the two business units, documents of understanding and all, with the ceremony of contracting an outside vendor, and moved the hosted business’s operations to that team. The offering relaunched under a new name with pricing that matched reality: steep, tiered list-price cuts, and a printed discount-ceiling schedule underneath so the new honesty could not be quietly discounted away. Alongside it we built pricing plays instead of pricing exceptions, my favorite being Step Up to SaaS, a declining-discount ramp that let a migrating customer carry the dual cost of old and new platforms without anyone shredding the margin to get the deal done. The escalations calmed, and the margins climbed from underwater to positive over the year that followed. The move freed two teams at once: the business unit’s own SaaS operations group, which had been fighting the hosted fires while also running our other cloud products, and the engineering lab, which could finally turn to the only work that would actually change our trajectory.
The mission transfer that brought back the mojo
The engineering heart of WebSphere Commerce was a lab in Toronto, full of people who had worked on the platform for their entire careers, and when I started visiting I found a team that everyone had written off, and that had half started to believe it. The talent was there. The mojo was not. I spent a lot of hours on planes trying to understand why, and the answer finally came from an engineer who leveled with me: it was impossible to be in innovation mode while perpetually repairing the legacy platform through customer escalations and support tickets. The unlock came from inside the lab’s own walls, not from me. There was already a division of labor with our sister lab in Beijing; what Toronto needed was a mission transfer, a real mechanism inside IBM, that would formally move the legacy platform’s care to Beijing and point Toronto entirely at the SaaS future; Beijing’s charter became protecting the base.
What I contributed was not the idea. It was the willingness to hear it, and the persistence to push it through the bureaucracy, and that is the whole lesson. Very quickly after, the magic started. Scaffolding appeared. Pieces began to land. The engineering leader at the center of it was Leho Nigul, who ran the technical transformation and later wrote up its philosophy publicly under a title I love, “Monolith to Pebbles”. The ambition had been set in writing before a line of the new code existed: in the space of one month in my first fall, the metric for activating a new customer changed in my own team plan from a number of days to a number of minutes. Development on the reimagined SaaS began in January 2017. First customers were live on it by early summer, under a new name chosen deliberately to retire two decades of baggage, IBM Digital Commerce, and by year end it was a production service running on two continents: call it twelve months from first commit to a SaaS selling in the US and Europe. By the time the offering matured we could provision twenty new customer environments in less time than one environment used to take. The days had, in fact, become minutes.
For calibration, when Oracle made a comparable move with its acquired commerce platform, an admittedly deeper multi-tenant rebuild, the journey took over four years. And our team was not whole for any of it: the freezes and reductions took nearly half the people who started the journey before it ended. That they shipped anyway is a fact about them, not about my process. Watching a team that had been counted out deliver at that speed remains one of the sweetest things I have been part of.
We were just as deliberate about how we sold it, because IBM had launched cloud commerce before and burned trust doing it. One of my strategy slides carried the title “‘No one gets fired for buying IBM.’ We must be better stewards of this iconic belief and earn every online store,” and we ran the launch that way: a written verification ladder that widened the aperture one complexity dimension at a time, from a US, English, B2C store on day one out through new currencies, tax regimes, payment methods, and geographies, each dimension proven with a real customer, then documented well enough that a partner could deliver the next one without our engineers in the room. Every release note opened by defining exactly the customer we could delight that day. Slower on paper, faster in truth: the platform never got ahead of its own promises again.
Boulders to pebbles
The base could not be left behind, so the same architecture went back into the flagship. In December 2017 we shipped WebSphere Commerce version 9, the most disruptive release in the platform’s modern history: the monolith decomposed toward services, delivered as Docker containers with immutable images, running on a lightweight application server, with a new externalized customization model that finally broke the trap my advisor had diagnosed before I joined. Because customizations now lived outside our code, customers could finally take our releases without treating each upgrade as a re-implementation project, which meant, for the first time in years, IBM could ship innovation and the base could actually absorb it. It is the same trap that catches any vendor whose best customers customize, in any industry, and the fix travels. The metaphor we took on the road was the one Leho’s team lived: from boulders to pebbles. And through all of it, the installed base held: the support revenue that funded the whole program stabilized during the transition, because nothing we shipped forced a customer to move before they were ready. We rebuilt the engine without stalling the aircraft.
The containers also forced a family argument. IBM’s own Power systems, a storied hardware franchise with a couple hundred of our commerce customers on it, did not yet support the container runtime the new architecture required, and I refused to pretend otherwise: the future of this product only ran where containers ran, and I said so, in writing, in rooms where it was unwelcome. A wise mentor inside IBM advised me to let the question escalate rather than settle it quietly, and he was right. By January 2018 the Power team had full Docker support for Linux and was offering us environments to smoke test it, and AIX was not on the list. Their engineers built that, not me; what I built was the argument that made the question impossible to leave unanswered.
Then we took the story to the base, because a platform whose customers had stopped watching needed to be re-introduced to its own future. I did a roadshow around the world, my favorite stop being my first trip to Australia, walking rooms of customers who had not seen meaningful change in years through what was coming and what it would ask of them. The job was not announcing features; it was re-selling a skeptical installed base on its own future. In January 2018 we launched v9 at the NRF show with a packed client breakfast, and the industry took the measure of it partly by who showed up: the keynote seats went to Accenture’s Brian Walker and the CIO of Yoox Net-a-Porter. The commercial turn had already begun: after we put the new strategy and offerings into the market in the third quarter of 2017, the pipeline that had gone quiet during the transition came back to life, and in our operating reviews, fourth-quarter signings grew year over year for the first time in the whole SaaS transition. The rocking chair was not a speedboat yet. But it was visibly, publicly, moving.
Where my AI story starts
Everything above is the turnaround. Threaded through it is the origin of the arc I have been on ever since IBM put the word Watson in my title.
The rebrand that gave my business unit its name was, to my surprise, the most rigorous branding exercise I have ever been inside of. To call a product Watson, it was not enough to gesture at algorithms; the product had to pass a real assessment against IBM’s written definition of a cognitive solution, capability by capability: did it understand, did it reason, did it learn, did it interact naturally. I was impressed then and I am more impressed now; the industry has still not adopted that discipline about the word AI, and it shows.
Inside that frame, my teams got to build things whose time had not yet come. In the fall of 2016 I ran a product effort we called conversational commerce: enterprise-grade virtual agents that could hold a multi-turn dialogue, understand intent, and guide a shopper to a purchase; reading the sentiment and tone of written language was planned next. Our own deck called the product a cognitive agent. Now, 2016 was the year everyone built a chatbot, so I will not claim we were alone on that wave. What set the work apart was the ambition: agents wired into catalog, pricing, and order systems, with real retail prospects lining up and one asking to be a development partner. What we lacked was a decade of technology that had not been invented yet.
The market now calls these AI agents and considers them the future of shopping. The term was closer than anyone knew: an internal roadmap from my first month described one prototype as a flexible AI agent framework, in exactly those words, even while everything customer-facing stayed carefully cognitive. In a detail I could not have invented, that roadmap lived on an internal site named for the wild ducks. We were early the way the whole wave was early, and we knew it even then. The timing lesson stuck: vision tells you where the puck is going; arithmetic tells you whether you can get there before your funding runs out.
The same conviction ran through the other two portfolios. In Merchandising, the DemandTec team had just delivered IBM Dynamic Pricing, a fully organic new product that watched market signals and recommended price moves in something close to real time; my era was building its business case and taking it to market. And in the analytics portfolio, the one furthest from the commerce story I have been telling, sat Coremetrics, the web-analytics pioneer IBM had acquired in 2010 and let age, and Tealeaf, which could replay a struggling customer’s session click by click. That portfolio ran on a doctrine the team called macro to micro, one I still reach for. Classic analytics keeps score, and a score tells you whether you are winning, never why. The goal was an analyst who could go from macro to micro in one motion: from the KPI that moved, to the machine-learned read on what was driving it, to the replay of a single human struggling with a single form field. We re-architected the aging Coremetrics onto a real-time platform so that descent could happen at modern data volumes; my Intel years had left me convinced the flood of connected devices would soon demand nothing less. The rebuilt platform earned its keep the hard way: it carried the holidays without a single severity-one incident and kept marquee clients from defecting. Its ceiling was set elsewhere. The company chose not to fund a frontal fight with Adobe and Google for the analytics market, a call I understood and still regret. Some fights inside a portfolio that big you win in the product and lose in the budget.
I also fed myself. I spent over a hundred hours of personal time that year formally studying machine learning, along with the Docker and JavaScript stacks my teams were adopting, because I have never trusted a product leader’s roadmap conviction that outruns their technical understanding, least of all my own.
The strategic acquirer’s seat
One more education IBM gave me that no smaller company could have: at IBM, you are the strategic acquirer, and corp dev is always sliding something across your desk. As the portfolio leader I evaluated a steady stream of it, and the two decisions I remember best were both recommendations against. We went deep on a CPQ partnership, technical due diligence and all, and I recommended we pass: the economics handed most of the value to the partner, the platform came with cloud lock-in we could not accept, and the strategic exposure of placing a third party between our commerce and order management franchises was worse than the gap we were filling. We built forward on our own CPQ instead. And when Magento crossed my desk, as most things in commerce eventually did if you sat where I sat, I recommended we pass on that too: we had rounded the corner on our own SaaS, and for the enterprise clients we served, I believed what we were building was the better platform. In May 2018, Adobe announced it was buying Magento for $1.68 billion, completing a strategy I had helped write there years earlier. One advisor had joked back in 2016 that joining IBM was fine, so long as I got it to buy Magento. This industry keeps its receipts.
The flash drive
IBM’s Think conference in March 2018 gave us the biggest stage the company owns, and I challenged Leho’s team to unveil IBM Digital Commerce there, not with slides but with a live provisioning run: a real customer would hand us their storefront content on a flash drive at the start of the session, and the team would stand up their running store before it ended, in front of the audience. The demo had one non-negotiable requirement: it had to be real. Real production systems, a real customer’s real data, nothing agreed in advance beyond a small catalog and its imagery, nothing that would not survive scrutiny from the WebSphere Commerce customers’ engineers in the seats. What I brought from Adobe was not the substance but the theatrics, the conviction that a true thing lands harder when you stage it like theater.
We committed to the theater completely; we wore Carhartt gear on stage, stocking cap included in my case. Steve Mello, my boss, opened the session, and he is the reason the tension in the room was real: he had championed his wild duck inside IBM, he knew exactly how big a deal it was to nail this in public, and you can hear both the confidence and the stakes in his voice as he explains what we are about to attempt. Carhartt’s e-commerce leader handed over the drive. The team went to work on the spot, the store came up, and the whole thirty-six minutes is on YouTube: a twenty-two-year-old platform’s brand-new SaaS edition, live, delivering in one conference session what its predecessor used to need weeks to do. It was also the release philosophy in miniature: we had designed IBM Digital Commerce to toggle new features on at industry events, NRF, Shoptalk, Think, while promising customers that their storefront was theirs and nothing shopper-facing would ever change on IBM’s clock instead of the customer’s. Days later, I walked out of IBM for the last time. I can think of worse closing arguments.
The ending
Through my last year, the mandate I had nearly quit waiting for, and then received, and then delivered, quietly drained away again. Growth plans were reset from far above us, budgets tightened every quarter, and the energy that had recruited an outsider class in 2016 was visibly redirecting toward other bets. By the time Elastic Path called about leading marketing for their Series B, I could read where IBM was heading with the portfolio: commerce was simply ceasing to be strategic to IBM, no matter what the products did. The mission I had been recruited for was ending. I had delivered what I had signed up to deliver, the reimagined SaaS was live and had been demonstrated live at Think, and I did not leave behind an empty chair: the 2018 plan had successors and promotions penciled in, and Platform 2.0 was already sketched for whoever came next. On March 31, 2018, I chose the Series B.
The epilogue played out in public. That December, IBM announced the sale of a family of software products, WebSphere Commerce among them, to HCL for $1.8 billion, and Forrester’s analysts wrote the sentence that still stops me: after the deal closed, IBM would no longer have a commerce platform. The following year the analytics and pricing assets I had run went to a private equity carve-out called Acoustic. HCL, to its credit, took the v9 we built and carried it forward; it ships to this day as HCL Commerce, which means the modernization that Toronto team pulled off has now outlived IBM’s ownership of it by seven years and counting. The product leader I had groomed to succeed me crossed over with the portfolio and still leads product there. The SaaS we raced to build went across in the same deal and then simply faded: HCL built its cloud future on the v9 codebase instead, and IBM Digital Commerce never got a public goodbye, just an end-of-support notice where its documentation used to live, which is its own kind of epitaph. Leho and I were both bummed for a long time about the ending we did not get to write. We had real working upgrades that could have changed that business, and we were, as the saying goes, a day late and a dollar short. The chorus of 2016 had called it a salvage job, not a resurrection, and a decade on I score it a draw: they were right about the corporation, and wrong about what that team could still build.
Two symmetries soften it. The first I only remembered while writing this. In college, the only company I pursued besides the consultancy that hired me was IBM, and specifically its e-commerce software team, the people building Net.Commerce, the direct ancestor of the platform this whole chapter is about. I am not sure they ever wrote back. Sixteen years later I was the IBM executive responsible for it. Some applications just take a while to process. The second is that within months of my departure, Leho and I had each landed at Elastic Path, independently and without coordinating, which tells you everything about what two people fresh from rebuilding one of the biggest commerce platforms in the world thought the next act would look like.
What I take from it
IBM is where I learned to operate at a scale nothing else in my career approaches, and where I learned that at that scale, the org chart is the product strategy. Almost every unlock in this chapter, the hosted business’s rescue, Toronto’s revival, even my own near-departure, turned on organizational mechanics rather than technology: moving a mission to the team built for it, formally relieving builders of the past so they could build the future, printing the impediments list and handing it to the one person who would read it straight. I walked in believing strategy documents change companies. I walked out knowing that at enterprise scale, strategy is what you rewire the organization to make inevitable, and I now read org charts the way other product people read roadmaps.
It is also where my pricing craft was forged, in gold text boxes and corrected margins, and where I learned the leadership move I now reach for first: when a team has gone flat, do not give a speech. Go visit, keep asking, and listen for the unlock, because the team usually already knows what it is, and what they need is not inspiration but an executive willing to spend political capital making it official. And it is where my conviction about AI in commerce started, early enough to be wrong about the timing, and to watch the industry, years later, arrive where that work had been pointing.
Mostly, though, IBM taught me what I am for. Kareem’s wild ducks stayed with me longer than any strategy we wrote. There is a kind of company that needs its untamed hires, and a kind of hire that needs to stay untamed, and the best thing leadership can do is hold that tension the way he did across a lunch table, when I arrived carrying a resignation dressed up as a list of impediments. I got my twenty months of flying in formation with a giant: waking its oldest platform, pricing like a shopkeeper, shipping like a startup, and leaving the estate younger than I found it. Then I went looking for the opposite problem. Buried in my earliest packaging plans, sketched in the fall of 2016 for a 2018 tier I never got to ship, was the whole platform unbundled into services a customer could buy à la carte: catalog, pricing, promotions, cart, each as a service. The industry would eventually call that composable commerce, and I had concluded I did not want to wait for a giant to get around to it. I had just revived one of the largest collections of commerce software in the industry; the next thing I wanted was the smallest company with the one piece that mattered, built the way the future was clearly going to demand. What an API-first sliver of a platform can do against the giants, I went to find out.
This chapter is part of My Work, my career told one company at a time.
← Intel · All chapters · Elastic Path →
Elsewhere from this era: IBM’s answer man (Retail Leader interview, 2017)