Top 10 in Tech - What to know for Week ending April 17, 2026

SaaS METRIC OF THE WEEK: SaaS METRIC OF THE WEEK: GES

Growth Endurance Scoreis a metric that assesses a company's ability to sustain growth over time (something I have been discussing quite a bit lately, trying to maintain my own growth momentum). GES measures this efficiency by factoring in both net retention and customer acquisition efficiency. A high GES correlates with long-term business health and resilience. This score provides valuable insights for businesses aiming for consistent, sustainable growth. Bessemer has drilled deeper into it and plotted ARR growth YoY, and found that the decay is fairly predictable at 30%. That's a benchmark - so in other words, you should expect next year's growth rate to be 70% of the current year, as the stakes get higher.

GROWTH ENDURANCE

TL:DR #1: this year's growth % divided by last year's - is the metric your investors are running but probably not telling you about. Public market median is 92%. In private markets, 80% is best in class. The math is pretty brutal tho: 70% endurance on 12% growth goes to 8%, then 6%, then 4% in three years. Dropbox is now in revenue decline. Palantir grew from 29% to 59%. Datadog has held above 25% at $3.4B in revenue. The difference shows up in the endurance score long before it shows up in the headline number.

DISCOUNT

I had a good convo about discounting and deals earlier this week - Discounting can juice growth but wreak havoc on your metrics. Both ChartMogul and SaaStr warn: reflect actual revenue in MRR, not list price. Misreporting MRR distorts GTM signals and investor trust. But you can also increase prices so salespeople can "discount".

RAISE

A whopper guide (149 pages) for all of y'all in raise mode right now - the guide provides actionable insights for navigating the complexities of raising capital, covering investor relations, pitching essentials, market awareness, and the fundraising process.

MARKETING

Why do some B2B SaaS ads actually land? Because they nail 3 truths: product (what you do), emotional (why it matters), and cultural (why now). Stripe and Slack get it. Most don't.

AI SPEAK

With AI, our tech dictionaries can get out of date fast! Check this Q1 2026 update that just flags the terms that actually changed - not because people got bored, but because the behavior changed. The sharp one: HITL now means a human approves specific agent actions before they happen, not just reviewing outputs. Guardrails went from output filters to a full governance stack covering who can deploy, what gets logged, and who gets paged.

TRUST

Great stat - 95% of AI pilots fail to deliver any measurable impact - not because the models are weaksauce but because the product never earns trust. Users judge AI on its worst moments, not its average. A confident wrong answer loses more credibility than an uncertain correct one - because we end up not knowing WHICH outputs we can trust. The tools people actually keep are the ones that show what they don't know, let users undo things, and keep human judgment (HITL - see #6 above) in the loop.

DATA

Here is a push back on the "95% of AI pilots fail" narrative above (#7) from a16z with some actual enterprise numbers (or maybe this is the 5%?): 29% of Fortune 500 and 18.5% of Global 2000 are AI live, paying customers of a leading AI startup - signed contracts, converted pilots, in production. Coding generates $3B in annualized startup revenue, an order of magnitude above everything else!!!!! Legal ($500M) and support ($400M) are the next clearest wins.

CRM

I've been having the CRM discussion my whole career, and the perpetual "Which CRM" decision is now shifting from features to agents. This SaaStr breakdown argues the winning platforms will be those with the strongest AI agent ecosystems, where automation, data orchestration, and workflow execution matter more than traditional UI or pipeline management.

CASE STUDY

AGENTS: Complementing 7 and 8 above, SaaStr went from 20+ employees to 3 humans plus 20 AI agents - and the revenue swung from -19% to +47% YoY, $4.8M in pipeline directly attributed to agents. Lemkin's honest post-mortem on why most implementations fail: bad data exposed immediately, scaling broken processes faster, generic messaging, no human owner, and quitting after month one. The number that lands hardest - it took 47 iterations to stop their AI SDR being too aggressive on pricing. Most teams quit in month two.