Small companies often add tools the way they add coffee orders: one more, because today feels busy. A one-page technology budget — categories, a monthly ceiling, named owners, and renewal dates — stops stack drift without killing useful experiments, including AI trials.
Start with categories, not brands
List buckets: core operations (site, email, payments), growth (ads, CRM, analytics), productivity (docs, chat, project tools), security/backup, and experiments (AI, prototypes). Brands change; buckets stay. When a vendor rebrands or bundles three products into one SKU, your map still makes sense.
If you sell online, compare this page to a full picture of what running an online business really costs — hosting, payment fees, and “small” SaaS lines that only look small alone.
Write the number you can afford
Pick a monthly ceiling as a share of revenue or a fixed euro amount. Everything new must fit under it or replace something else. “We will grow into it” is how stacks bloat. A ceiling also forces trade-offs: ads versus another analytics suite, a CRM upgrade versus an AI trial.
Tag each tool: must, useful, habit
Must — business stops without it.
Useful — clear time or money saved.
Habit — someone liked the demo once.
Kill or downgrade habits first. Habit tools are where how digital subscriptions quietly add up hide: overlapping plans, unused seats, annual renewals nobody calendared.
Plan renewals and owners
Every line gets a person responsible and a renewal date. No orphan subscriptions. Review 30 days before auto-renew, not the morning the card declines. Owners also decide who may add a card — shadow purchases are a budget leak dressed as initiative.
Price AI like a line item, not a miracle
AI spend shows up as seats, API usage, and “copilot” add-ons. Treat tokens and usage caps explicitly — see counting AI token costs alongside headcount — instead of hoping the invoice stays polite. Ask the same question you would for any automation: whether automation actually lowers company costs and whether automation really pays off. If you cannot name the workflow and the metric, it belongs in the experiment slice, not in core ops.
Leave a thin experiment line
Reserve a small slice for trials — including AI — with an end date. If it does not earn a “useful” tag in 30–60 days, cut it. A technology budget is not anti-innovation; it is anti-amnesia.
When the experiment is “build instead of buy,” keep whether SaaS faces a crisis as teams build their own apps and build costs on the same page so the comparison stays honest.
Security is not optional plumbing
Backup, access control, and basic hardening belong in must — not in “we will do it after the launch.” Cheap tools that skip security create expensive days. Use a short list of common cybersecurity mistakes in small companies as a filter before you approve another login that touches customer data.
Quarterly ritual (45 minutes)
1) Export card and app-store charges.
2) Map each charge to a bucket and tag.
3) Cut or downgrade habits.
4) Reconfirm owners and renewals.
5) Reallocate savings to must-fixes or the experiment line.
Do this four times a year and your stack stops being a museum of last year’s demos. For bigger bets, use the same scepticism you would for checking whether an AI investment really pays off: tools earn their keep, or they leave.
How to set the ceiling without a finance degree
Three workable methods: (1) a fixed euro amount that still hurts a little if wasted; (2) a percentage of revenue (often 1–4% for tool-heavy small firms, adjusted to your margin); (3) last quarter’s spend minus an intentional cut target. Pick one. Revisit when headcount or revenue jumps by more than ~20%.
Separate capital-ish items (laptops, one-time migrations) from operating subscriptions so a hardware refresh does not “prove” that SaaS is cheap.
Unit economics for tools
Translate spend into units you feel: cost per active employee, cost per order, cost per published article, cost per closed deal. When a tool cannot attach to a unit, it is either infrastructure (keep, but name it) or vibes (cut).
AI lines should show both seat cost and usage cost. If usage is spiky, set soft alerts before the invoice becomes a post-mortem. That is the operational twin of counting AI token costs alongside headcount.
Negotiation and downgrade playbook
Before renewal: export usage, list unused seats, ask for a downgrade path in writing. Vendors expect silence; a calm email with numbers often yields a better tier or multi-month credit. Bundle renewals where one owner can trade seats across products.
If a product is strategic, still ask checking whether an AI investment really pays off questions yearly. Strategic is a claim, not a permanent exemption from arithmetic.
Example one-page budget layout
Columns: bucket, tool, monthly cost, owner, renewal date, tag (must/useful/habit), note. Footer: ceiling, current total, gap, experiment remaining. Keep it in a shared doc. Screenshot it into the quarterly meeting. Beauty is optional; continuity is not.
When online costs confuse the picture, reconcile with what running an online business really costs so ads, cloud, and SaaS do not hide in different spreadsheets.
What “good” looks like after six months
Put risk on the same sheet as tools: one mistake, one client, and a 100,000 PLN liability bill can wipe a careful budget overnight.
Fewer orphan tools. Named owners. Experiment line that actually turns over. No surprise renewals. A written no-list (“we will not buy another design suite this year”). Security basics funded. AI spend tied to workflows, not FOMO. That is a technology budget doing its job.
Brak komentarzy:
Prześlij komentarz