/articlesIT operations
IT operations through conversation: triage, investigate, report
AI for IT operations with Botify: it reads Dynatrace, Splunk and ServiceNow to triage alerts, find what changed and write the incident report, with approval.
- Updated
- Published
- min read
- 06
- On this page
- 08
On this page08
- What does Botify do in IT operations?
- How is Botify different from classic AIOps?
- Which IT-operations tasks does Botify handle?
- What does an investigation look like end to end?
- Which actions should Botify be allowed to take?
- Where does Botify go wrong in IT operations?
- How to start with Botify in IT operations
- Common questions
01
What does Botify do in IT operations?
Botify answers your engineers’ incident questions from the tools they already use. It reads a question or an alert, picks the system to ask, reads the result and decides what to ask next. Today it works with Dynatrace and Splunk for monitoring and ServiceNow for tickets and changes.
Its answers come from your systems, not from memory. A general chatbot can explain what a Dynatrace problem is. Botify tells you which problems are open on your checkout service right now and what changed before they started, with the record behind each statement.
Think of it as an on-call helper that never tires of copying IDs between tabs. It collects the evidence, puts it in order and hands your senior engineer a short summary to decide on.
02
How is Botify different from classic AIOps?
AIOps means AI for IT operations. Classic AIOps tools cut alert noise: they take in large volumes of events and metrics, learn what normal looks like, spot anomalies and group related alerts into one incident. That job still matters, and those tools do it well.
Botify works one step later. It does not replace anomaly detection; it starts from its output. A Dynatrace problem or a Splunk ITSI notable event is where Botify begins. It then pulls deployments, logs, change records and past incidents from the other tools the AIOps engine never sees.
| Classic AIOps | Botify | |
|---|---|---|
| Main job | Spot anomalies, cut noise, group events | Answer questions and investigate across tools |
| Input | Streams of metrics, events and logs in one platform | A question, an alert or an incident, plus access to several tools |
| Output | Alerts, problems, anomaly scores | A written answer or report that cites its evidence |
| A new kind of question | Needs a new rule, model or dashboard | Answered if the right tools are connected |
| Taking action | Runbooks set up in advance | Proposes an action; a person approves before it runs |
| Typical failure | Noisy or missed alerts when normal behaviour shifts | A confident answer without evidence, if sources are not required |
03
Which IT-operations tasks does Botify handle?
Four jobs, and each is mostly reading and summarising. That makes them safe to start with.
- Triage: read a new Dynatrace problem, check ServiceNow for a related incident, look at recent changes to the service, then say how serious it looks and which team probably owns it.
- Investigation: collect problems, deployments, logs, traces and change records for one service and one time window, then rank the likely causes with the evidence for each.
- Reports: turn the investigation into a shift handover or an incident summary as a PDF, Excel, Word or CSV file, in Arabic or English.
- Proposed actions: draft the incident update or the deployment marker, then wait until a person approves it.
04
What does an investigation look like end to end?
Here is how Botify handles “Why is checkout slow since 10:00?” with the actions it has today. The order is not scripted: Botify picks the next step from what the last one returned.
- List open Dynatrace problems for the last two hours and read the one on the checkout service, with its root-cause evidence and affected services.
- List Dynatrace deployment events for the same window, and ServiceNow changes planned near that time.
- Run a Splunk search for errors on the checkout hosts in the same window, with a time limit and a result cap, and check for ITSI notable events.
- Match the service across tools, because Dynatrace, Splunk and ServiceNow rarely use the same name.
- Check which changes happened just before the slowdown, allowing a few minutes for clocks that disagree.
- Write a short answer: the most likely cause, the evidence for it, what would rule it out and a next step that waits for approval.
05
Which actions should Botify be allowed to take?
Sort every action by what it can change. Reading a metric is harmless. Recording a deployment event or updating an incident is not. Most teams start with Botify read-only and add write actions one at a time, each with approval.
In Botify, every action has a risk level (read, write, destructive or sends outside) and an approval setting. Opening or updating a ServiceNow incident and recording a Dynatrace deployment event always wait for a person. Every action and approval is kept in the audit log.
| Kind of action | Example | Sensible default |
|---|---|---|
| Read | Query metrics, search logs, list problems or change records | Runs straight away |
| Write | Update an incident state, record a deployment event | A person approves |
| Destructive | Delete, restart, roll back | Not connected at first; approval if ever turned on |
| Sends outside | Email someone outside the team from Gmail | A person approves |
06
Where does Botify go wrong in IT operations?
The mistakes are predictable, and each has a fix in how you set Botify up. A better prompt alone fixes none of them.
- Invented causes: the model fills a gap with a likely story. Require every finding to cite a tool result, and label anything without one as a guess to test.
- Runaway searches: an open-ended log search is slow and costly. Botify’s Splunk search refuses to run without a time window, and caps the results.
- Different names: “checkout”, “checkout-svc-prod” and a ServiceNow item may be the same thing. Keep a list of matching names across tools, and let a person confirm each new match.
- Timing taken as cause: a deployment near the incident is a lead, not a verdict. The Google SRE book describes troubleshooting as forming guesses and testing them, and that applies to assistants too.
- Too much access: Botify can do whatever the connected account can do. Grant read access by default and add write access on purpose.
07
How to start with Botify in IT operations
Start small. Pick one service with good monitoring and an on-call team willing to compare Botify’s answers with their own. Connect Dynatrace, Splunk or ServiceNow with read access only, and list the ten questions your engineers ask most during incidents.
Judge Botify on whether its answers are right and whether every claim leads back to a record, not on how well it writes. Once the team trusts the read-only answers, turn on one write action with approval, such as updating the incident, and read the audit log after each incident.
08
Common questions
01Does Botify replace AIOps platforms?
No. AIOps platforms spot anomalies and group events. Botify starts from what they found and checks the other tools. Most teams keep their AIOps platform and add Botify next to it.
02Can Botify fix incidents automatically?
It can propose a fix and prepare the action, but any change waits for a person to approve. Fully automatic fixes are better left to tested runbooks with a rollback.
03What access does Botify need for IT operations?
Read access to the tools your engineers use during incidents: Dynatrace or Splunk for problems, metrics, logs, traces and deployments, and ServiceNow for incidents, changes and configuration items. It does not keep its own copy of your monitoring data.
04How do you stop Botify from inventing a root cause?
Make evidence required: every finding points to a specific tool result, and anything without one is labelled as a guess. Your engineers can open the query or record behind each line.
05Is it safe to connect Botify to production monitoring?
Read-only access is low risk when the connected account has narrow permissions and searches are time-bound. The risk is in write actions, so keep them off at first and require approval when you turn them on.
09
Sources
10